Abstract
BACKGROUND & AIMS:
Heterogeneous presentations and disease mechanisms among patients with laryngeal symptoms account for misdiagnosis of laryngopharyngeal reflux (LPR), variations in testing and suboptimal outcomes. We aimed to derive phenotypes of patients with laryngeal symptoms based on clinical and physiologic data and to compare characteristics across phenotypes.
METHODS:
302 adult patients with chronic laryngeal symptoms were prospectively enrolled at three centers between 1/2018–10/2020 [age 57.2±15.2 years; 30% males; BMI 27.2±6.0 kg/m2]. Discriminant analysis of principal components (DAPC) was applied to 12 clinical and 11 physiologic variables collected in stable condition to derive phenotypic groups.
RESULTS:
DAPC identified five groups, with significant differences across symptoms, hiatal hernia size, and number of reflux events (p<0.01). Group A had the greatest hiatal hernia size (3.1cm±1.0;p<0.001) and reflux events (37.5±51;p<0.001), with frequent cough, laryngeal symptoms, heartburn and regurgitation. Group B had the highest body mass index (28.2kg/m2±4.6;p<0.001) and salivary pepsin (150ng/ml±157;p=0.03), with frequent cough, laryngeal symptoms, globus, heartburn and regurgitation. Group C frequently reported laryngeal symptoms (93%;p<0.001), and had fewest esophageal symptoms (9.6%;p<0.001) and reflux events (10.7±11.0;p<0.001). Group D commonly reported cough (88%;p<0.001) and heartburn. Group E (18%) was oldest (62.9y±14.3;p<0.001) and distinguished by highest integrated relaxation pressure.
CONCLUSIONS:
DAPC identified distinct clinico-physiologic phenotypes of patients with laryngeal symptoms referred for reflux evaluation: Group A, LPR and gastroesophageal reflux disease (GERD) with hiatal hernia; Group B, Mild LPR/GERD; Group C, No LPR/No GERD; Group D, Reflex cough; Group E, Mixed/Possible obstructive esophago-gastric junction. Phenotypic differences may inform targeted clinical trials design and improve outcomes.
Keywords: Gastro-esophageal reflux disease, Esophageal manometry, Proton pump inhibitor
INTRODUCTION
One-third of the adult population experiences laryngeal complaints such as throat clearing, dysphonia, or chronic cough. Laryngeal symptoms can arise from various etiologies including allergy, post-nasal drip, vocal cord dysfunction, visceral hypersensitivity, or laryngopharyngeal reflux (LPR). LPR is an extra-esophageal syndrome of gastroesophageal reflux disease (GERD) in which retrograde reflux of gastric contents leads to chronic laryngeal irritation.1–3 While laryngeal symptoms are not synonymous with LPR, up to 80% of symptomatic individuals seeking consultation may receive a diagnosis of LPR with the vast majority treated with empiric proton pump inhibitor (PPI) therapy.4–7 Furthermore, LPR is often diagnosed in the absence of typical esophageal symptoms of GERD.1 These empirical one-size-fits-all approaches for heterogeneous non-specific laryngeal complaints drive variations observed in physiologic testing and suboptimal efficacy of conventional medical and surgical anti-reflux therapies.3,8 Thus, methods to delineate mechanisms of symptomatology and clarify whether GERD is a culprit are critically needed.
Recent efforts to phenotype patients with reflux symptoms based on clinical characteristics and physiologic profiles have been described to allow distinct and personalized management in GERD.9 Consequently, the paradigm is shifting away from the historic approach that largely focused on empiric acid suppression, and moving towards a phenotype-driven, personalized approach to GERD. Similarly, unsupervised clustering for identification of patient subpopulations have been successfully applied to other disease states such as heart failure, cancer and diabetes.10 These approaches are needed among patients with laryngeal symptoms to understand symptom pathogenesis, appropriately diagnose LPR, and formulate tailored, personalized management approaches. Therefore, we aimed to cluster patients with laryngeal symptoms referred for reflux evaluation based on a range of clinical and physiologic data using unsupervised phenotype modeling and compare characteristics across phenotypes.
METHODS
Study Design & Subjects
This multi-center study analyzed prospectively collected data between January 2018 to October 2020 from patients with laryngeal symptoms. The study was approved by the Institutional Review Board at participating sites and data share agreements were executed (University of California San Diego, La Jolla, CA; University of Colorado, Aurora, CO; Brigham and Women’s Hospital, Boston, MA).
Adults presenting with at least 8 weeks of one or more extra-esophageal symptom (dysphonia, sore throat, throat clearing, cough, or globus), with or without concomitant esophageal symptoms, who underwent esophageal high-resolution impedance-manometry (HRIM) were enrolled. Patients with history of prior foregut or otolaryngolic surgery, foregut or otolaryngologic malignancy, chronic lung disease and eosinophilic esophagitis were excluded.
Data Collection & Management
The prospective dataset included 302 subjects with 12 clinical parameters (n=302), 4 HRIM metrics (n=302), 6 ambulatory reflux monitoring metrics (n=251), and 1 salivary biomarker value (n=77). Clinical parameters included demographics, presenting symptoms, use of PPI, and validated patient-reported symptom instruments.
HRIM was performed using a 36-channel solid-state catheter (Manoscan 360; Medtronic, Minneapolis, Minnesota; or Diversatek, Highlands Ranch, Colorado) and a standardized protocol of ten supine 5ml wet swallows. Metrics collected included upper esophageal sphincter (UES) resting pressure, median lower esophageal sphincter (LES) integrated relaxation pressure (IRP), manometric hiatal hernia size (separation between crural diaphragm and LES), % of ineffective swallows, and % of complete bolus clearance. Validated symptom instruments included GerdQ11 and reflux symptom index (RSI)12 scores.
Ambulatory reflux monitoring performed included MII-pH monitoring (Reflux 6.1; Medtronic, Minneapolis, Minnesota; or Diversatek, Highlands Ranch, Colorado) or wireless pH monitoring (Bravo; Medtronic, Minneapolis, Minnesota). MII-pH catheter configurations included single-pH electrode in the distal esophagus, or dual-pH electrodes in the distal esophagus and at the UES. Patients that underwent ambulatory reflux monitoring were instructed to follow their baseline regular diet during monitoring. Data collected from ambulatory reflux monitoring included study type, on/off acid suppression, total acid exposure time (% time pH below 4.0), number of reflux events, number of full column reflux events, number of LPR events, and distal mean nocturnal baseline impedance (DMNBI). The pH-impedance studies were manually reviewed, with reflux events defined by at least 50% drop in impedance from baseline and propagating in retrograde fashion from the distal most electrode pairs. LPR reflux events were defined by an impedance reflux event reaching the most proximal electrode pairs. Unstimulated fasting salivary pepsin was collected, processed within seven days of collection, and quantified for salivary pepsin concentration using the Peptest lateral flow device (LFD) (RD Biomed Ltd) per previously published protocols.13 14
Phenotype-Based Clustering Using Discriminant Analysis of Principal Components
The primary outcome was phenotypic cluster grouping based on discriminant analysis of principal components (DAPC), an unsupervised multivariate method involving principal components analysis, K-means clustering and discriminant analysis to detect and visualize clusters of related individuals across a population structure.15 The steps to conduct the DAPC analysis are outlined in a tutorial, which we followed and briefly outline here.15 First, an a priori selection of variables to include in the model was identified, motivated by potential clinical utility and to minimize missingness in the available data while maximize sample size. The variables included in the primary model included age, body mass index (BMI), hiatal hernia size, normalized LES IRP, the percent of ineffective swallows (distal contractile integral < 450 mmHg-s-cm), and presence of symptoms related to cough, throat symptoms (a composite of voice hoarseness, sore throat, or throat clearing), globus, regurgitation, heartburn and chest pain. Second, k-means clustering was used to identify between 3 to 6 initial clusters, where this range was a priori selected to maintain a reasonable number of clusters to evaluate for potential clinical utility. Third, the k-means clusters were then used in the DAPC, where the optimal number of principal components was selected based on the a-score when peaked or a lower value when plateaued to avoid over-fitting the data, with the number of discriminant functions retained based on the change in eigenvalues for adding additional functions. Finally, this “optimal” DAPC model was used to assign group membership based on the individual probability to generate figures, summarize performance, and propose potential phenotypes. Evaluation of the DAPC analyses for the different group sizes was conducted to select the number or groups that balanced the overlap of group membership while providing clinically meaningful differences.
To evaluate if detailed ambulatory reflux monitoring can provide fine mapping of patients beyond the clinical and HRIM parameters, a subgroup analysis of subjects that had available ambulatory reflux monitoring data was conducted to evaluate potential group membership. To evaluate if clinical presentation alone could reliably identify phenotypes, a sensitivity analysis was conducted to evaluate potential group membership based on symptoms alone in the model. Cramér’s V was calculated to estimate the association between the primary grouping with the subgroup and sensitivity groups.
Comparison of variables across DAPC phenotypes used an overall F-test for an unadjusted linear regression model for continuous measures and a chi-squared test for categorical measures. All categorical measures were summarized as count (percent) and all continuous measures as mean (standard deviation). An adjusted median LES IRP was calculated by dividing the observed value with the established normative cutoff (15 mmHg for Medtronic and 20 mmHg for Diversatek) to normalize across the two systems. All analyses were conducted in R v3.6.3 (Vienna, Austria) and DAPC was implemented using the adegenet package17 and Cramér’s V was implemented using the DescTools package.15
RESULTS
In total, 302 subjects were enrolled [90 (30%) males, mean age of 57.2 years (SD 15.2), mean BMI of 27.2 kg/m2 (SD 6.0)]. The predominant extra-esophageal symptom reported was throat clearing (237, 78%), followed by dysphonia (172, 57%), globus (160, 53%), cough (159, 52%), and sore throat (45, 15%). Typical esophageal symptoms of GERD reported included 110 (36%) with heartburn, 93 (31%) with regurgitation, and 47 (16%) with non-cardiac chest pain.
Primary Outcome: Phenotypic Cluster Groups
DAPC analysis of the complete data set (n=302) identified five distinct groups: 26 (9%) group A, 72 (24%) group B, 94 (31%) group C, 57 (19%) group D, and 53 (17%) group E (Figure 1A). Probability of a subject being correctly assigned to each group were high for groups A, B, and E (95%, 81%, and 77%, respectively) and lower for groups C and D (68% and 55%, respectively) (Figures 1B & 1C). Of all included variables the loading variables significantly influencing cluster groups were hiatal hernia size, median IRP, and presenting symptoms (Figure 2).
Figure 1. Five Phenotype Cluster Groups.
1A)Group A, B, E do not overlap with the other five groups; groups C and D have minimal overlap. 1B)Waterfall plot of individual probabilities for five groups. X-axis represents a patient; y-axis represents membership probability. 1C)Boxplot of posterior probability for each group based on group membership
Figure 2. Loading Variables Driving Group Membership.
The total contribution of variable to loading sums to 1.0 within each loading (1 and 2). This barplot is a graphical representation of loading variables that had a measure >0.001, and reflect a relative measure of the contribution of each original variable to the discriminant functions of the DAPC.
Distribution of Variables Across Phenotypic Cluster Groups (Figure 3)
Figure 3.
Distribution of Variables Within Each Group.
Notable differences among cluster groups are summarized below (p-values reflecting comparisons across all five groups) and displayed in Table 1.
Table 1.
Distribution of Clinical and Physiologic Variables Across Phenotype Clusters
| Phenotype Clusters | ||||||
|---|---|---|---|---|---|---|
| LPR/GERD hiatal hernia | Mild LPR/GERD | No LPR/No GERD | Reflex Cough | Mixed/Possible Obstructive EGJ | ||
| A (n=26) | B (n=72) | C (n=94) | D (n=57) | E (n=53) | p-value | |
| Demographics | ||||||
| Age (years) | 53.6 (14.0) | 52.2 (13.3) | 58.8 (15.5) | 56.5 (16.3) | 62.8 (14.3) | <0.001 |
| Male | 8/26 (30.8%) | 20/72 (27.8%) | 29/94 (30.9%) | 19/57 (33.3%) | 13/53 (24.5%) | 0.87 |
| BMI (kg/m2) | 27.4 (4.6) | 28.2 (6.1) | 26.8 (6.4) | 26.8 (5.1) | 26.4 (6.6) | 0.46 |
| Symptom Presentation | ||||||
| Cough | 22/26 (84.6%) | 48/72 (66.7%) | 10/94 (10.6%) | 50/57 (87.7%) | 28/53 (52.8%) | <0.001 |
| Laryngeal Symptoms | 18/26 (69.2%) | 71/72 (98.6%) | 87/94 (92.6%) | 48/57 (84.2%) | 11/53 (20.8%) | <0.001 |
| Voice Hoarseness | 11/26 (42.3%) | 54/72 (75.0%) | 68/94 (72.3%) | 32/57 (56.1%) | 7/53 (13.2%) | <0.001 |
| Sore Throat | 4/26 (15.4%) | 22/72 (30.6%) | 4/94 (4.3%) | 12/57 (21.1%) | 3/53 (5.7%) | <0.001 |
| Throat Clearing | 16/26 (61.5%) | 66/72 (91.7%) | 82/94 (87.2%) | 43/57 (75.4%) | 7/53 (13.2%) | <0.001 |
| Globus | 13/26 (50.0%) | 58/72 (80.6%) | 64/94 (68.1%) | 22/57 (38.6%) | 1/53 (1.9%) | <0.001 |
| Regurgitation | 15/26 (57.7%) | 59/72 (81.9%) | 2/94 (2.1%) | 14/57 (24.6%) | 2/53 (3.8%) | <0.001 |
| Heartburn | 12/26 (46.2%) | 60/72 (83.3%) | 9/94 (9.6%) | 23/57 (40.4%) | 5/53 (9.4%) | <0.001 |
| Chest Pain | 5/26 (19.2%) | 30/72 (41.7%) | 4/94 (4.3%) | 7/57 (12.3%) | 1/53 (1.9%) | <0.001 |
| High Resolution Manometry | ||||||
| Hiatal Hernia Size (cm) | 3.1 (1.0) | 0.3 (0.6) | 0.0 (0.0) | 0.2 (0.5) | 0.3 (0.7) | <0.001 |
| Hernia Present | 25 (96.2%) | 17 (23.6%) | 2 (2.1%) | 12 (21.1%) | 11 (20.8%) | <0.001 |
| Normalized LES IRP Ratio | 0.5 (0.3) | 0.6 (0.3) | 0.6 (0.3) | 0.7 (0.3) | 0.9 (0.6) | <0.001 |
| Median LES IRP (Medtronic) (mmHg) [N=110] | 6.8 (5.0) | 10.7 (5.0) | 12.6 (7.2) | 13.3 (5.4) | 20.0 (9.1) | <0.001 |
| Median LES IRP (Diversatek) (mmHg) [N=192] | 9.2 (6.4) | 11.1 (4.7) | 11.9 (5.5) | 10.7 (4.8) | 11.9 (5.5) | 0.48 |
| Percent ineffective swallows | 45.7 (32.3) | 22.9 (30.9) | 28.2 (26.7) | 28.6 (31.7) | 32.3 (35.3) | 0.03 |
| Mean UES Basal Pressure (mmHg) [N=206] | 74.8 (38.1) | 89.4 (64.1) | 99.3 (50.6) | 91.0 (57.2) | 105.0 (69.7) | 0.42 |
| Ambulatory Reflux Monitoring | ||||||
| Number of Reflux Events [N=251] | 37.6 (51.4) | 19.2 (25.9) | 10.7 (11.0) | 26.2 (27.2) | 23.4 (18.4) | <0.001 |
| Number of LPR Events [N=187] | 7.2 (17.4) | 2.4 (3.6) | 4.8 (7.4) | 2.3 (4.1) | 3.0 (5.8) | 0.13 |
| Overall acid exposure time (%) [N=258] | 2.6 (4.1) | 1.5 (4.1) | 1.4 (2.4) | 1.6 (2.6) | 3.9 (6.9) | 0.01 |
| Acid exposure time off PPI (%) [N=204] | 1.9 (1.9) | 1.6 (4.4) | 1.5 (2.6) | 1.5 (2.5) | 4.2 (8.0) | 0.05 |
| Acid exposure time on PPI (%) [N=54] | 3.8 (6.2) | 0.8 (2.2) | 0.5 (1.1) | 2.0 (3.1) | 3.3 (2.7) | 0.09 |
| DMNBI (Ohms) [N=226] | 1.8 (1.3) | 2.8 (1.3) | 2.2 (1.2) | 2.5 (1.2) | 2.1 (1.0) | 0.01 |
| Symptom index | 56.5 (33.5) | 41.5 (39.2) | 29.4 (36.3) | 39.1 (29.0) | 34.8 (19.7) | 0.61 |
| Positive SAP for extra-esophageal symptom (>95%) | 3/7 (42.9%) | 14/24 (58.3%) | 3/6 (50.0%) | 14/22 (63.6%) | 5/11 (45.5%) | 0.80 |
| Other Variables | ||||||
| Salivary pepsin concentration (ng/ml) [N=77] | 107.2 (116.7) | 150.5 (156.6) | 77.7 (85.7) | 68.2 (88.3) | 40.1 (40.8) | 0.03 |
| GerdQ Score [N=213] | 8.9 (3.2) | 8.5 (2.7) | 7.0 (2.6) | 7.6 (2.9) | 7.1 (3.0) | 0.01 |
| RSI Score [N=258] | 24.8 (12.8) | 22.7 (8.9) | 17.8 (9.0) | 20.5 (9.0) | 16.8 (13.1) | <0.01 |
Body mass index (BMI); Lower esophageal sphincter (LES); Integrated relaxation pressure (IRP); Upper esophageal sphincter (UES); Laryngopharyngeal reflux (LPR); Proton pump inhibitor (PPI); Distal mean nocturnal baseline impedance (DMNBI); Symptom association probability (SAP); Reflux symptom index (RSI); Gastroesophageal reflux disease (GERD); Esophagogastric junction (EGJ).
Demographics:
Group B was youngest (52.2 years (SD 13.3)) and group E was oldest (62.9 (14.3)) (p<0.001). Groups A and B had higher BMI (27.4 kg/m2 (SD 4.6) and 28.2 (6.1)) compared to Groups C, D, and E (26.8 (6.4), 26.8 (5.1), 26.4 (6.6)), respectively.
Presenting Symptoms:
With regards to extra-esophageal symptoms, cough was reported by the majority in Groups A (85%), B (67%), and D (88%) (p<0.001). Laryngeal symptoms (dysphonia, throat clearing, and/or sore throat) were reported by the majority in Groups A (69%), B (99%), C (93%), and D (84%) (p<0.001). Globus was reported by the majority in Groups B (81%) and C (68%) (p<0.001). With regards to esophageal symptoms, regurgitation was reported by the majority in group A (58%) and B (82%) (p<0.001), and heartburn in group B (83%) (p<0.001).
High Resolution Manometry:
Hiatal hernia size was greatest in Group A (3.1cm (SD 1.0) and smallest in Group C (0.0 (SD (0.0); p<0.001). Median IRP was greatest in Group E (normalized ratio 0.9 (0.6); p<0.001). Group A exhibited the greatest proportion of ineffective swallows (46% (SD 32); p=0.03). For UES metrics, Group A exhibited the lowest UES basal (74.8 mmHg (38)) and residual (0.7 (7.5)) pressures.
Other Variables:
Ambulatory reflux monitoring data were available for 252 (83%) patients. Group A exhibited the greatest number of total (37.6 (SD 51); p<0.001) and LPR (7.2 (17.4); p=0.13) reflux events. Mean DMNBI was lowest in Group A (1790 ohms (SD 1340; p=0.01). Acid exposure time on PPI was highest in Group A (3.8% (SD 6.2); p=0.09) and off PPI was highest in group E (4.2% (SD 8.0); p=0.05). Mean salivary pepsin concentrations were highest in Group B (150 ng/ml (SD 157)) and Group A (107 (SD 117); p=0.03).
Post-hoc Clinical Characterization of Phenotypic Cluster Groups
In order to provide clinical context, we labeled the five cluster groups based on the results of the DAPC of clinico-physiologic characteristics as follows: LPR/GERD with hiatal hernia (Group A), mild LPR/GERD (Group B), no LPR/GERD (Group C), reflex cough (Group D), and mixed with possible obstructive EGJ process (Group E). (Figure 4)
Figure 4. Representative Depiction of Phenotypic Cluster Groups.
Five distinct clusters depicted based on symptom presentation (left bullets) and clinical/physiologic profiles (right arrows) with arrow representing directionality and degree (bolding) of relationship. Group A: Commonly presented with hiatal hernia, ineffective esophageal peristalsis, cough, laryngeal symptoms, heartburn and regurgitation; higher BMI, AET on PPI, number of reflux and LPR events, salivary pepsin; lowest DMNBI. Group B: Presented with highest BMI and salivary pepsin concentration; commonly presented with cough, laryngeal symptoms, globus, heartburn and regurgitation. Group C: Lowest proportion with hiatal hernia and number of reflux events; commonly presented with laryngeal symptoms and globus, however rarely with esophageal symptoms. Group D: highest proportion of patients with cough and commonly presented with heartburn. Group E: presented with the highest mean LES IRP, and predominantly with symptoms of cough; lowest salivary pepsin concentration.
Comprehensive Sub-Group Model Inclusive of Symptom Presentation, High Resolution Manometry, & Ambulatory Reflux Monitoring Table 2
Table 2.
Distribution of Clinical and Physiologic Variables Across Clusters in Sub-Group Analysis
| SubGroup Clusters | |||
|---|---|---|---|
| Sub Group A (n=9) | Sub Group B (n=64) | Sub Group C (n=178) | |
| Demographics | |||
| Age (years) | 45.4 (7.39) | 43.4 (13.8) | 63.0 (12.3) |
| Male | 3 (33.3%) | 21 (32.8%) | 53 (29.8%) |
| BMI (kg/m2) | 28.0 (3.98) | 26.4 (6.42) | 27.1 (5.92) |
| Symptom Presentation | |||
| Cough | 7 (77.8%) | 39 (60.9%) | 84 (47.2%) |
| Laryngeal Symptoms | 7 (77.8%) | 50 (78.1%) | 149 (83.7%) |
| Voice Hoarseness | 4 (44.4%) | 34 (53.1%) | 114 (64.0%) |
| Sore Throat | 4 (44.4%) | 17 (26.6%) | 13 (7.3%) |
| Throat Clearing | 6 (66.7%) | 43 (67.2%) | 143 (80.3%) |
| Globus | 5 (55.6%) | 30 (46.9%) | 106 (59.6%) |
| Regurgitation | 4 (44.4%) | 26 (40.6%) | 38 (21.3%) |
| Heartburn | 3 (33.3%) | 28 (43.8%) | 57 (32.0%) |
| Chest Pain | 2 (22.2%) | 18 (28.1%) | 22 (12.4%) |
| High Resolution Manometry | |||
| LES-CD separation; Hiatal Hernia Size (cm) | 1.4 (1.4) | 0.4 (0.8) | 0.3 (0.9) |
| Hernia Present | 5 (55.6%) | 11 (17.2%) | 21 (11.8%) |
| Normalized LES IRP Ratio | 0.6 (0.6) | 0.7 (0.4) | 0.6 (0.3) |
| Percent ineffective swallows | 45.6 (40.6) | 7.73 (11.5) | 38.1 (31.3) |
| Mean UES Basal Pressure (mmHg) | 58.2 (10.2) | 84.9 (47.3) | 102 (63.5) |
| Ambulatory Reflux Monitoring | |||
| Number of Reflux Events | 112 (47.3) | 29.3 (20.4) | 11.8 (11.5) |
| Overall acid exposure time (%) | 3.97 (5.66) | 1.89 (4.1) | 1.94 (4.1) |
| Acid exposure time off PPI (%) | 3.96 (4.29) | 1.98 (4.66) | 1.96 (4.29) |
| Acid exposure time on PPI (%) | 3.99 (7.81) | 1.64 (1.91) | 1.85 (3.18) |
| DMNBI (Ohms) | 2.02 (1.36) | 2.7 (1.38) | 2.26 (1.15) |
| Positive SAP for extra-esophageal symptom (>95%) | 6 (66.7%) | 31 (62.0%) | 60 (51.3%) |
| Other Variables | |||
| Salivary pepsin concentration (ng/ml) | 106 (125.3) | 77.3 (100.9) | 43.5 (71.1) |
| GerdQ Score | 9.4 (2.61) | 7.75 (2.93) | 7.59 (2.77) |
| RSI Score | 25.5 (11.9) | 19.3 (9.56) | 19.9 (9.98) |
Body mass index (BMI); Lower esophageal sphincter (LES); Integrated relaxation pressure (IRP); Upper esophageal sphincter (UES); Laryngopharyngeal reflux (LPR); Proton pump inhibitor (PPI); Distal mean nocturnal baseline impedance (DMNBI); Symptom association probability (SAP); Reflux symptom index (RSI); Gastroesophageal reflux disease (GERD); Esophagogastric junction (EGJ).
DAPC sub-group analysis limited to subjects who underwent ambulatory reflux monitoring (n=251) identified three cluster groups (Cramér’s V of 0.21 (95% CI: 0.07 to 0.27) with the primary analysis, Supplemental Table). Sub-Group A (n=9) was reminiscent of an LPR/GERD with hiatal hernia phenotype characterized by greatest hiatal hernia, presentation with regurgitation, BMI, salivary pepsin concentration, acid exposure time, number of reflux events, and lowest DNMBI. 0% of sub-group A had membership in the original Group C (no LPR/no GERD). Sub-Group B (n=64) was reminiscent of mild GERD and commonly presented with cough, laryngeal symptoms and esophageal symptoms, and had a moderate salivary pepsin concentration. Patients in Sub-Group B most frequently had membership in the primary Groups B (mild LPR/GERD) (33%) and D (cough reflex) (25%). Sub-Group C (n=178) was characterized by frequent cough and laryngeal symptoms, smallest hiatal hernia size and the lowest salivary pepsin concentration. Sub-Group C (reminiscent of no LPR and no GERD) had rare membership in the primary Group A (LPR/GERD with hiatal hernia; 11/178, 6%).
Sensitivity Analysis: Symptoms Only Model
DAPC analysis including only symptoms (n=302) identified five cluster groups, with a Cramér’s V of 0.51 (95% CI: 0.44, 0.56) with the primary analysis (Supplemental Table). Of note, of the 94 patients originally assigned to the primary Group C (no LPR/no GERD) phenotype clustering based on symptoms alone correctly assigned 75 (80%) into a cluster with rare esophageal symptoms.
DISCUSSION
Assessment of patients with laryngeal symptoms referred for reflux evaluation has remained challenging for clinicians and investigators. Patients may present with a variety of throat symptoms that mimic other otorhinolaryngologic conditions, often without typical esophageal symptoms of GERD.2 While empiric acid suppression therapy is often the first-line treatment, the response is variable16 and prior studies on the efficacy for surgical intervention for suspected LPR have been inconclusive.3 These variations illustrate the significant heterogeneity of patients with laryngeal symptoms suspected to have LPR. Effective management would, therefore, require careful selection of patients and therapies tailored to the underlying mechanisms. This first-of-its-kind study applied sophisticated phenotyping methods to explore potential clinical physiologic phenotypes of patients with laryngeal symptoms undergoing evaluation for reflux.
This study of 302 adults presenting with laryngeal symptoms who underwent HRIM at three centers identified five distinct clinico-physiologic phenotypes. Groups A and B likely represents a spectrum of GERD with esophago-pharyngeal reflux. Notably a small proportion (8.6%) had membership in Group A (LPR/GERD with hiatal hernia), a group characterized by disrupted anti-reflux barrier and higher reflux burden. Group B (LPR/mild GERD) shares characteristics with group A, though less often has a hiatal hernia and more often has positive symptom-reflux association, suggesting a milder spectrum of GERD with potential of reflux hypersensitivity overlap. On the other hand, Group C (No LPR/No GERD) likely represents a non-GERD pathology with isolated laryngeal symptom presentation and least suggestion of reflux physiology. Of note membership was most prevalent for Group C (30.9%). Groups D and E are less well characterized and in sensitivity analyses share overlap between phenotypes resembling GERD as well as those without GERD (Supplemental Table). Group D (Reflex cough) may be comprised of patients whose laryngeal symptoms (cough) were in part reflux-related, perhaps through the described vagally-mediated reflex mechanism leading to bronchoconstriction, rather than esophago-pharyngeal reflux.17,18 Group E (Mixed/possible obstructive EGJ) may include patients with an obstructive physiology across the EGJ, which may lead to poor bolus clearance, chronic esophageal distention, and increased pressurization in the proximal esophagus. It is important to note that an elevated supine median IRP on manometry does not equate to LES dysfunction, and the elevated IRP may be related to catheter artifact or effect of hiatal hernia.
Our results of this proof-of-concept study illustrate the value of systematic assessment through clinico-physiologic phenotyping and personalized management, particularly prior to escalated anti-reflux therapy. We applied the results from this study to develop a post-hoc classification tree model in which 81% of patients were classified in the correct phenotype. Literature to-date has not identified any single diagnostic test as being adequately predictive for an accurate diagnosis of LPR and treatment outcome. Given the current lack of a “gold standard” diagnosis for LPR, tailored management based on clinical and physiologic phenotyping using a multitude of data may provide a cost-effective and efficient approach to these challenging patients with laryngeal symptoms.
Using results from our phenotyping, potential clinical algorithms may be formulated for a sensible tailored management approach. In a potential model generated from our current data (Figure 5), assessment for esophageal symptoms (heartburn or regurgitation), BMI, or presence of hiatal hernia may serve as the initial step to dichotomize patients into the GERD groups A (LPR/GERD + Hiatal hernia) or B (LPR/mild GERD) versus groups C (No LPR/No GERD), D (Reflex Cough) or E (Mixed/Possible obstructive EGJ). Notably, our sensitivity model based solely on symptom presentation correctly assigned 80% of patients into the no LPR/noGERD phenotype (Group C), supporting a pre-test probability that patients presenting with laryngeal symptoms in the absence of typical esophageal symptoms likely do not have LPR. This potential diagnostic approach to laryngeal symptoms echoes concepts in the clinical model by Patel et al., which identified that the overall risk of pathologic reflux among patients with refractory esophageal or extraesophageal symptoms was independently associated with heartburn, elevated BMI, asthma and presence of hiatal hernia.20
Figure 5. Conceptual Diagram of Clinical Approach to Identifying Phenotypes of Patients with Suspected LPR.
This diagram synthesizes results from this study and clinical experiences. The first step is assessment of symptoms, BMI, and presence of hiatal hernia. Patients with high BMI and concomitant heartburn/regurgitation likely fall into GERD groups A (LPR/GERD with hiatal hernia) or B (LPR/mild GERD), and testing including ambulatory reflux monitoring should be performed for further characterization. Those without heartburn/regurgitation likely belong to group C (no LPR/no GERD) or possibly group D (reflex cough) or E (Mixed/Possible obstructive EGJ). Evaluation can include upper gastrointestinal endoscopy with real-time functional luminal impedance-planimetry, esophageal HRM, and/or barium esophagram. In absence of obstructive physiology, ambulatory reflux monitoring should be performed to further stratify patients into groups with GERD (groups A, B, or potentially D) versus those without GERD (group C).
After this initial step of dichotomization, evaluation strategies may then follow to allow efficient phenotyping and tailored therapy. Patients with LPR and pathologic GERD with hiatal hernia (group A) would be the optimal candidates for escalated anti-reflux therapy, including potential surgical intervention.8 Patients with LPR and mild reflux or reflux hypersensitivity (group B) and those with reflux-induced reflex cough (group D) may benefit from “mild” anti-reflux management with acid suppression or alginates.21 Invasive interventions such as anti-reflux surgery should generally be avoided due to unclear benefits in these groups. Neuromodulation or behavioral interventions may be particularly beneficial for patients in group B with reflux hypersensitivity.22 Gabapentin, other neuromodulators, and voice therapy for cough suppression, desensitization, and avoidance strategies may be useful for group D patients with reflex cough.23 Patients with no LPR/no GERD (group C) warrant assessment and management of other etiologies, such as laryngeal abnormalities, vocal cord dysfunction, allergies/post-nasal drip, oropharyngeal dysphagia, or airway hypersensitivity. Finally, a sub-group with confirmed obstructive physiology at the EGJ on supportive testing may benefit from LES directed therapy. Further research is needed to understand response to treatment and outcomes across distinct phenotypes.
Our exploratory study has several strengths, including the multi-center cohorts, prospective data collection, and relatively large sample size for a study of LPR patients. There are also limitations to our study. Treatment response was not assessed, as this proof-of-concept study was designed as the first-of-its-kind to identify distinct clinical and physiologic phenotypes. A next step will be to validate this analysis incorporating outcome data. Modality of ambulatory reflux monitoring was not standardized and for this reason, ambulatory reflux monitoring data were not included in the primary DAPC model, but rather in the sensitivity analysis. Nonetheless, given the lack of a “gold standard” test and clear consensus of the optimal ambulatory reflux monitoring for evaluating LPR, and the limited availability of some of these newer technologies, the inclusion of all available types of ambulatory reflux monitoring likely improved the generalizability of our results. This study included the well-established extra-esophageal symptoms such as dysphonia, sore throat, throat clearing, globus, cough, however symptoms that may potentially suggest LPR such as halitosis or odynophagia were not within inclusion criteria. Finally, the included patients came from three tertiary care institutions with subspecialists in LPR that were referred to HRM which may include more severe or refractory patients compared to the general population. However, these refractory patients also likely represent the diagnostically challenging patients most frequently referred to gastroenterologist offices for evaluation, and would, therefore, most benefit from a tailored management approach.
In conclusion, distinct clinical physiologic phenotypes were identified in a multi-center prospective study utilizing DAPC among 302 patients with laryngeal symptoms referred for reflux evaluation. This exploratory study affirms and highlights that patients with suspected LPR are heterogeneous and a reflux physiology is not always present. Distinct clinical presentations and physiologic differences may explain variability in presentation, treatment response, and prognosis. Identifying phenotypic differences may improve outcomes and inform future targeted clinical trials.
Supplementary Material
WHAT YOU NEED TO KNOW.
Background:
Laryngopharyngeal reflux is commonly suspected with poorly delineated evaluation and management pathways
Patients with laryngeal symptoms suspected to have LPR are heterogeneous with variation in mechanisms of symptoms, presence of gastro-esophageal reflux, and treatment outcome.
Findings:
Discriminant analysis of principal components across 302 patients with chronic laryngeal symptoms referred for reflux evaluation identified five distinct phenotypic groups described as LPR and gastroesophageal reflux disease (GERD) with hiatal hernia (Group A); Mild LPR/GERD (Group B); No LPR/No GERD (Group C); Reflex cough (Group D); Mixed/Possible obstructive esophago-gastric junction (Group E).
Implications for patient care:
Distinct groups of patients with laryngeal symptoms referred for reflux evaluation exist and can be identified through clinical physiologic phenotypes
Phenotypic differences may inform targeted clinical trials design and improve outcomes
Research Funding Support:
American College of Gastroenterology Junior Faculty Development Award (PI: Yadlapati)
Relevant Conflicts of Interest:
Rena Yadlapati: Institutional Consulting Agreement: Medtronic, Ironwood Pharmaceuticals, Diversatek; Consultant: Phathom Pharmaceuticals; Research support: Ironwood Pharmaceuticals; Advisory Board with Stock Options: RJS Mediagnostix
John E. Pandolfino: Consultant: Medtronic, Ironwood Pharmaceuticals, Diversatek; Research support: Ironwood Pharmaceuticals, Takeda; Advisory Board: Medtronic, Diversatek; Stock Options: Crospon Inc
None: AKM, DRS, MG, JXC, TLC, SG, SW, TCW, PMK, PW, AMVH, WWC
Abbreviations:
- LPR
laryngopharyngeal reflux
- DAPC
discriminant analysis of principal components
- BMI
body mass index
- RSI
reflux symptom index
- GERD
gastroesophageal reflux disease
- PPIs
proton pump inhibitors
- HRIM
high-resolution impedance-manometry
- UES
upper esophageal sphincter
- LES
lower esophageal sphincter
- IRP
integrated relaxation pressure
- DMNBI
distal mean nocturnal baseline impedance
- LFD
lateral flow device
Footnotes
Transcript Profiling: None
Writing Assistance: None
REFERENCES
- 1.Koufman JA. Laryngopharyngeal reflux is different from classic gastroesophageal reflux disease. Ear Nose Throat J 2002;81(9 Suppl 2):7–9. (https://www.ncbi.nlm.nih.gov/pubmed/12353431). [PubMed] [Google Scholar]
- 2.Lechien JR, Bobin F, Muls V, et al. Gastroesophageal reflux in laryngopharyngeal reflux patients: Clinical features and therapeutic response. Laryngoscope 2020;130(8):E479–E489. DOI: 10.1002/lary.28482. [DOI] [PubMed] [Google Scholar]
- 3.Lechien JR, Dapri G, Dequanter D, et al. Surgical Treatment for Laryngopharyngeal Reflux Disease: A Systematic Review. JAMA Otolaryngol Head Neck Surg 2019;145(7):655–666. DOI: 10.1001/jamaoto.2019.0315. [DOI] [PubMed] [Google Scholar]
- 4.Vakil N, van Zanten SV, Kahrilas P, Dent J, Jones R, Global Consensus G. The Montreal definition and classification of gastroesophageal reflux disease: a global evidence-based consensus. Am J Gastroenterol 2006;101(8):1900–20; quiz 1943. DOI: 10.1111/j.1572-0241.2006.00630.x. [DOI] [PubMed] [Google Scholar]
- 5.Peery AF, Crockett SD, Barritt AS, et al. Burden of Gastrointestinal, Liver, and Pancreatic Diseases in the United States. Gastroenterology 2015;149(7):1731–1741 e3. DOI: 10.1053/j.gastro.2015.08.045. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Altman KW, Stephens RM, Lyttle CS, Weiss KB. Changing impact of gastroesophageal reflux in medical and otolaryngology practice. Laryngoscope 2005;115(7):1145–53. DOI: 10.1097/01.MLG.0000165464.75164.E5. [DOI] [PubMed] [Google Scholar]
- 7.Patel AK, Mildenhall NR, Kim W, Carroll TL. Symptom overlap between laryngopharyngeal reflux and glottic insufficiency in vocal fold atrophy patients. Ann Otol Rhinol Laryngol 2014;123(4):265–70. DOI: 10.1177/0003489414525021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Lechien JR, Saussez S, Schindler A, et al. Clinical outcomes of laryngopharyngeal reflux treatment: A systematic review and meta-analysis. Laryngoscope 2019;129(5):1174–1187. DOI: 10.1002/lary.27591. [DOI] [PubMed] [Google Scholar]
- 9.Yadlapati R, Pandolfino JE. Personalized Approach in the Work-up and Management of Gastroesophageal Reflux Disease. Gastrointest Endosc Clin N Am 2020;30(2):227–238. DOI: 10.1016/j.giec.2019.12.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Hedman AK, Hage C, Sharma A, et al. Identification of novel pheno-groups in heart failure with preserved ejection fraction using machine learning. Heart 2020;106(5):342–349. DOI: 10.1136/heartjnl-2019-315481. [DOI] [PubMed] [Google Scholar]
- 11.Jones R, Junghard O, Dent J, et al. Development of the GerdQ, a tool for the diagnosis and management of gastro-oesophageal reflux disease in primary care. Aliment Pharmacol Ther 2009;30(10):1030–8. DOI: 10.1111/j.1365-2036.2009.04142.x. [DOI] [PubMed] [Google Scholar]
- 12.Belafsky PC, Postma GN, Koufman JA. Validity and reliability of the reflux symptom index (RSI). J Voice 2002;16(2):274–7. DOI: 10.1016/s0892-1997(02)00097-8. [DOI] [PubMed] [Google Scholar]
- 13.Hayat JO, Gabieta-Somnez S, Yazaki E, et al. Pepsin in saliva for the diagnosis of gastro-oesophageal reflux disease. Gut 2015;64(3):373–80. DOI: 10.1136/gutjnl-2014-307049. [DOI] [PubMed] [Google Scholar]
- 14.Yadlapati R, Kaizer A, Greytak M, Ezekewe E, Simon V, Wani S. Diagnostic performance of salivary pepsin for gastroesophageal reflux disease. Dis Esophagus 2020. DOI: 10.1093/dote/doaa117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.A S. DescTools: Tools for descriptive statistics. . R package version 09934 2020. [Google Scholar]
- 16.Guo H, Ma H, Wang J. Proton Pump Inhibitor Therapy for the Treatment of Laryngopharyngeal Reflux: A Meta-Analysis of Randomized Controlled Trials. J Clin Gastroenterol 2016;50(4):295–300. DOI: 10.1097/MCG.0000000000000324. [DOI] [PubMed] [Google Scholar]
- 17.Mansfield LE, Hameister HH, Spaulding HS, Smith NJ, Glab N. The role of the vague nerve in airway narrowing caused by intraesophageal hydrochloric acid provocation and esophageal distention. Ann Allergy 1981;47(6):431–4. (https://www.ncbi.nlm.nih.gov/pubmed/7325414). [PubMed] [Google Scholar]
- 18.Schan CA, Harding SM, Haile JM, Bradley LA, Richter JE. Gastroesophageal reflux-induced bronchoconstriction. An intraesophageal acid infusion study using state-of-the-art technology. Chest 1994;106(3):731–7. DOI: 10.1378/chest.106.3.731. [DOI] [PubMed] [Google Scholar]
- 19.Patel DA, Sharda R, Choksi YA, et al. Model to Select On-Therapy vs Off-Therapy Tests for Patients With Refractory Esophageal or Extraesophageal Symptoms. Gastroenterology 2018;155(6):1729–1740 e1. DOI: 10.1053/j.gastro.2018.08.038. [DOI] [PubMed] [Google Scholar]
- 20.McGlashan JA, Johnstone LM, Sykes J, Strugala V, Dettmar PW. The value of a liquid alginate suspension (Gaviscon Advance) in the management of laryngopharyngeal reflux. Eur Arch Otorhinolaryngol 2009;266(2):243–51. DOI: 10.1007/s00405-008-0708-7. [DOI] [PubMed] [Google Scholar]
- 21.Sawada A, Guzman M, Nikaki K, et al. Identification of Different Phenotypes of Esophageal Reflux Hypersensitivity and Implications for Treatment. Clin Gastroenterol Hepatol 2020. DOI: 10.1016/j.cgh.2020.03.063. [DOI] [PubMed] [Google Scholar]
- 22.Dong R, Xu X, Yu L, et al. Randomised clinical trial: gabapentin vs baclofen in the treatment of suspected refractory gastro-oesophageal reflux-induced chronic cough. Aliment Pharmacol Ther 2019;49(6):714–722. DOI: 10.1111/apt.15169. [DOI] [PubMed] [Google Scholar]
- 23.Cohen SM, Misono S. Use of specific neuromodulators in the treatment of chronic, idiopathic cough: a systematic review. Otolaryngol Head Neck Surg 2013;148(3):374–82. DOI: 10.1177/0194599812471817. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.