Long-term Continued Proton Pump Inhibitor Use is Common in Patients Diagnosed with Eosinophilic Esophagitis Despite Failure of Histologic Response: Data from a two-centre study

Abstract

Background and Aims:

Treatment paradigms for proton pump inhibitors (PPIs) in eosinophilic esophagitis (EoE) are evolving. We aimed to determine patterns of long-term PPI use after EoE diagnosis in PPI histologic non-responders.

Methods:

We conducted a retrospective review at University of Colorado (UCH) and University of North Carolina (UNC) of EoE patients who were histologic non-responders to PPIs. Data were extracted from electronic medical records related to demographics, PPI use, and reasons for continuing or stopping PPI.

Results:

Of 67 patients in the UCH cohort, PPIs were initially discontinued in 9 (13%). Of 58 remaining on PPI, 48% were not instructed to discontinue therapy and 26% continued for symptom improvement. Of 675 patients at UNC, PPI was stopped in 185 (27%). Of patients remaining on PPI, 15% were not told to discontinue therapy and 62% were continued for symptom improvement. At last contact, >50% of patients remained on PPI at both centres with most common reasons for continuation being symptom improvement and not telling patients to discontinue. In the UNC cohort, clinical features associated with remaining on PPI included children younger than 18 years (p=0.01), males (p<0.001), heartburn symptoms (p<0.001) and hiatal hernia (p=0.004). Patients with dysphagia were less likely to remain on PPIs (p<0.001).

Conclusions:

Long-term PPI use is common in EoE patients even without histologic response. Failure to instruct patients to discontinue therapy was a common reason for long-term use, thus PPI use should be revisited in all EoE patients to confirm clinical benefit.

Introduction:

The criteria for diagnosis of eosinophilic esophagitis (EoE) have evolved since the first consensus guidelines in 2007,¹ with the most recent iteration published in 2018.² The previous diagnostic guidelines required symptoms of oesophageal dysfunction and persistence of oesophageal eosinophilia >15 eosinophils per high-power field (eos/hpf) after a 2-month proton pump inhibitor (PPI) trial.^{3, 4} The guidelines were originally rooted in the assumption that gastroesophageal reflux disease (GERD) and EoE were mutually exclusive, and that PPI-responsive oesophageal eosinophilia (PPI-REE) was a separate entity from EoE.⁴ However, this paradigm was repeatedly questioned,⁵ and new data informed updated international diagnostic criteria.² Current consensus considers PPIs as a treatment for EoE and not part of the diagnostic criteria, and patients with EoE who respond to PPIs (formerly PPI-REE) should now be considered as part of the EoE spectrum.^{2, 6}

While our understanding of the effect of PPIs in EoE has evolved, the role of PPIs in the treatment of EoE remains an active area of investigation. Independent of their effect on gastric acid secretion, PPIs have been shown to reduce oesophageal eosinophilia through other anti-inflammatory pathways,^{7, 8} and these mechanisms effectively treat the histologic, clinical, and endoscopic features of EoE.⁹ As such, they now stand as one of the potential first line therapies for EoE. In this new climate, however, there is an important need for ongoing evaluation of the benefit of PPIs, particularly with recent investigations on sequelae of long-term PPI use.⁴ Clinicians must determine whether patients with EoE should be continued on PPI therapy or whether they should undergo evaluation without PPI.¹⁰ As prior to the 2018 guidelines a PPI trial was required for EoE diagnosis, it has not been clear how many PPI non-responsive EoE patients remain on PPI treatment.

The aim of this study was to determine the patterns of PPI use after diagnosis of EoE, particularly in the PPI non-responsive (“classically defined”) EoE population. We hypothesized that despite histologic non-response, continued PPI use was common.

Materials/Methods

We conducted a two-centre retrospective cohort study of EoE patients diagnosed at the University of Colorado (UCH) and University of North Carolina (UNC). The study was approved by the institutional review boards of each institution. Data were collected from 2013–2017 and included subjects 18 years and older who met 2013 American consensus guidelines for diagnosis of EoE.^4,3 Standard clinical care for patients suspected of EoE included an index endoscopy, 6–8 biopsies from the upper and lower oesophagus, and initiation of any PPI at 20–40 mg twice daily if biopsies revealed oesophageal eosinophilia ≥ 15/high power field (hpf). Endoscopy was repeated 8 weeks after twice daily (high-dose) PPI use, and EoE was diagnosed if eosinophilia > 15/hpf persisted.⁴ Individuals who did not undergo a PPI trial before diagnosis of EoE, did not have follow-up after diagnosis of EoE, or were not on PPI at the time of the initial clinic visit were excluded.

Using a standardized data collection form, we extracted data for all identified cases from electronic medical records related to demographics and PPI use. PPI use (yes/no) was specifically assessed at 1) the first follow-up communication and/or clinic visit after diagnostic EGD, and 2) the last point of patient contact in the electronic medical record. The medical record was also assessed for all potential reasons for continuation or stopping of PPI. For analysis, characteristics of the study population were summarized with descriptive statistics. Proportions of patients remaining on PPIs after their initial follow-up and at last contact were calculated for each study site. A bivariate analysis was conducted for each cohort to compare EoE cases who remained on PPI with those who stopped PPI. Means were compared with 2 sample t-tests and proportions were compared with chi square. Then, multivariable logistic regression was performed to assess for independent predictors of remaining on PPI therapy. For model construction, we initially included all variables that had p<0.20 on the bivariate analysis, and then sequentially removed variables from the model that did not have p<0.05 and these remained out of the model if there was not a change in estimate of >10% with their removal.

Results:

67 adult subjects were included from the UCH cohort (mean age 37 yrs; 60% male; 97% white) and 702 patients of any age with their PPI status known at the last contact in the system were included from the UNC cohort (mean age 26 yrs; 43% <18 yrs; 67% male; 81% white) (Table 1). All EoE subjects were confirmed to be histologic non-responders to the high-dose PPI trial and had data related to being on or off PPI at the last contact in the system.

Table 1.

Baseline Demographics

	University of North Carolina (n=702)	University of Colorado (n=67)
Age, mean years (SD)	26.3 (18.5)	37 (10)
Children < 18 years (n, %)	303 (43)	0
Male (n, %)	468 (67)	40 (60)
Caucasian (n, %)	560 (81)	65 (97)
Mean days of follow-up (SD)	953 (1048)	226 (270)
Median days of follow-up (IQR)	572 (152–1455)	131 (52–322)

SD: Standard deviation; IQR: Interquartile range

PPI use after failed PPI trial

Of 67 patients at UCH with PPI use data, PPI was discontinued after histologic failure in 9 subjects (13%). Of the 58 patients remaining on PPI, 28 (48%) were not told to discontinue therapy despite no documentation of benefit or indication for ongoing use. In the remaining 30 patients where PPI was continued, 26% were continued for symptom improvement, 9% were continued for improvement in histology, and smaller proportions were continued for histology, symptom, or endoscopic benefit (Table 2).

Table 2.

PPI use and indication after diagnostic EGD

	University of North Carolina	University of Colorado
Number remaining on PPI after initial trial	463 (94)	58 (87)
Reasons for remaining on PPI initially (n, %)
Did not tell patient to stop	70 (15)	28 (48)
Histology benefit	14 (3)	5 (9)
Symptom benefit	288 (62)	15 (26)
Endoscopy benefit	0 (0)	0 (0)
Histology & symptom benefit	29 (6)	3 (5)
Histology & endoscopy benefit	6 (1)	0 (0)
Symptom and endoscopy benefit	8 (2)	3 (5)
Symptom, endoscopy, & histology benefit	19 (4)	2 (3)
Study protocol requires continuation	20 (4)	0 (0)
Weaning off medication started	9 (2)	2 (3)

PPI: Proton pump inhibitor

Of 675 patients with data regarding PPI use after the initial PPI trial in the UNC cohort, PPI was initially stopped in 185 (27%) of patients after failure of the PPI trial. Of the 490 cases remaining on PPI, reasons for continuation were available in 463 (94%). These reasons were variable. For example, 15% of patients were not told to discontinue PPI, 62% were continued for symptom improvement, and, similar to the UCH cohort, small proportions were continued for either histology, symptom, endoscopic benefit, or some combination of the three (Table 2).

PPI use at last contact

In the UCH cohort, mean patient follow-up was 226 days. Of 50 patients remaining on PPI at the most recent patient encounter, continuation was due to symptom benefit (32%), histology benefit (6%), symptom and endoscopy benefit (2%), and symptom, endoscopy, and histology benefit (4%) (Table 3). In 20 patients (40%), patient was not told to discontinue PPI therapy.

Table 3.

PPI use and indication at the time of last patient contact

	University of North Carolina	University of Colorado
Mean patient follow-up (days)	953	226
Number remaining on PPI at last contact	396 (54%)	50 (75)
Reasons for remaining on PPI long-term (n, %)†
Did not tell patient to stop	45 (12)	20 (40)
Histology benefit	10 (3)	3 (6)
Symptom benefit	240 (66)	16 (32)
Endoscopy benefit	1 (0)	0 (0)
Histology & symptom benefit	29 (8)	2 (4)
Histology & endoscopy benefit	6 (2)	0 (0)
Symptom and endoscopy benefit	5 (1)	1 (2)
Symptom, endoscopy, & histology benefit	17 (5)	2 (4)
Study protocol requires continuation	6 (2)	0 (0)
Weaning off medication started	4 (1)	4 (8)

^†Data available for 363 of the subjects with information regarding reasons for remaining on PPI

PPI: Proton pump inhibitor

Mean patient follow-up in the UNC cohort was 953 days. When reviewing the last available patient encounters in the UNC cohort, 396 of 702 patients (56%) remained on PPI, and data regarding a reason for PPI use was known in 363 (92%). Continuation of PPI was related to symptom benefit (66%), histology and symptom benefit (8%), histology and endoscopy benefit (2%), and several other reasons (Table 3). There were 45 patients (12%) who were not told to discontinue PPI use despite no documentation of benefit or ongoing indication.

Predictors of maintaining PPI use at the last contact

In the UNC cohort where full data were available, there were multiple clinical features that were associated with remaining on PPI (n=396) compared to being off PPI (n=306) (Table 4). For example, children younger than 18 (47% vs 39%; p=0.01), males (72% vs 59%; p<0.001), patients with baseline heartburn symptoms (49% vs 29%; p<0.001), and patients with a hiatal hernia (13 vs 7%; p=0.004) were more likely to remain on PPIs. In contrast, patients with dysphagia (66% vs 78%; p<0.001) were less likely to remain on PPIs, and in this group typical endoscopic signs of EoE were also more common (Table 4). On multivariable logistic regression, female sex (OR 1.96; 95% CI: 1.40–2.73), presence of dysphagia (1.87; 1.31–2.68), lack of heartburn (1.63; 1.19–2.24), presence of endoscopic oedema (1.49; 1.07–2.06), and lack of hiatal hernia (2.69; 1.52–4.76) were all independently associated with remaining off PPI after histologic non-response to this treatment.

Table 4:

Features of patients remaining on and off PPI therapy (University of North Carolina cohort)

	Remain on PPI (n = 396)	Off PPI (n = 306)	p
Age at diagnosis (mean years ± SD)	25.5 ± 19.0	27.5 ± 17.7	0.16
Child < 18 years (n, %)	187 (47)	113 (38)	0.01
Male (n, %)	286 (72)	182 (59)	< 0.001
White (n, %)	317 (81)	243 (80)	0.95
Initial symptoms (n, %)
Dysphagia	259 (66)	238 (78)	< 0.001
Food impaction	106 (27)	110 (36)	0.03
Heartburn/reflux	179 (49)	88 (29)	< 0.001
Chest pain	41 (11)	34 (11)	0.94
Abdominal pain	93 (24)	53 (18)	0.14
Nausea	35 (9)	38 (13)	0.31
Vomiting	101 (26)	76 (25)	0.95
Failure to thrive	44 (11)	24 (8)	0.33
Baseline endoscopy findings (post-PPI)
Rings	166 (42)	161 (53)	0.005
Stricture	72 (18)	81 (27)	0.008
Narrowing	50 (13)	67 (22)	0.001
Furrows	230 (58)	208 (68)	0.007
Exudates	143 (36)	132 (43)	0.06
Oedema	126 (32)	129 (42)	0.005
Crepe-paper mucosa	19 (5)	12 (4)	0.58
Erosive esophagitis	70 (18)	53 (17)	0.91
Hiatal hernia	52 (13)	20 (7)	0.004
Dilation performed	78 (20)	77 (25)	0.09
Baseline peak eosinophil count (post-PPI; mean eosinophils per high-power field ± SD	62.9 ± 43.3	72.2 ± 47.4	0.009

PPI: Proton pump inhibitor; SD: Standard deviation

In the UCH cohort, there were fewer associations for those remaining on PPI (n=50) compared to those off PPI (n=17) (Table 5). Crepe-paper mucosa was more common in those off PPI than on PPI (24% vs 4%; p=0.02), but this was not a common finding. There were non-significant trends showing that heartburn and chest pain were more numerically more common in those on PPI while vomiting was less common. On multivariable logistic regression, symptoms of vomiting (OR 4.38; 95% CI: 1.21–15.9) and crepe-paper mucosa (OR 15.0; 95% CI: 1.61–140.3) were independently associated with remaining off PPI after histologic non-response, but estimates were not precise.

Table 5:

Features of patients remaining on and off PPI therapy (University of Colorado cohort)

	Remain on PPI (n = 50)	Off PPI (n = 17)	p
Age at diagnosis (mean years ± SD)	37.9 ± 10.9	35.3 ± 6.3	0.35
Child < 18 years (n, %)	0	0	--
Male (n, %)	30 (60)	10 (59)	0.73
White (n, %)	48 (98%)	16 (94%)	0.09
Initial symptoms (n, %)
Dysphagia	49 (98)	16 (94)	0.42
Food impaction	29 (58)	8 (47)	0.43
Heartburn/reflux	30 (60)	7 (41)	0.18
Chest pain	11 (22)	1 (6)	0.13
Abdominal pain	93 (24)	53 (18)	0.40
Nausea	3 (6)	1 (6)	0.99
Vomiting	16 (32)	10 (59)	0.05
Baseline endoscopy findings (post-PPI)
Rings	41 (82)	14 (82)	0.97
Stricture	17 (34)	7 (41)	0.59
Narrowing	7 (14)	1 (6)	0.37
Furrows	35 (70)	15 (88)	0.14
Exudates	14 (28)	4 (24)	0.72
Oedema	26 (52)	9 (53)	0.95
Crepe-paper mucosa	2 (4)	4 (24)	0.02
Erosive esophagitis	0	0	--
Hiatal hernia	8 (16)	2 (12)	0.67
Dilation performed	19 (38)	7 (41)	0.82
Baseline peak eosinophil count (post-PPI; mean eosinophils per high-power field ± SD	45.9 ± 24.3	36.5 ± 14.8	0.14

PPI: Proton pump inhibitor; SD: Standard deviation

Discussion

PPIs have been important in EoE diagnostic algorithms for more than a decade, and with the updated international consensus guideline, PPIs have been reclassified as a potential first-line therapy for EoE.^{2, 8, 11, 12} Implications of this guideline change may be that PPIs will be prescribed more frequently as monotherapy, concomitantly with conventional therapies such as topical steroids or diet elimination,¹³ or even prescribed at higher doses. Indeed, in a retrospective multicentre cohort study where patients with “classic” PPI-REE with initial histologic response to PPIs developed recurrent eosinophilia, 90% of relapsers were re-induced into histologic remission with dose increase of PPI from once to twice daily.¹⁴

Prior standards of care for diagnosis of EoE initially assessed whether or not patients were PPI-responsive, and for those who were PPI responsive, PPI treatment was typically continued as effects were felt to be maintained.¹⁴ For those who were histologic non-responders to PPI treatment, however, PPIs should have conceivably been stopped with initiation of other treatments in the absence of clinical benefit. However, the frequency with which PPIs were stopped in the face of non-response was unknown. In our study, attention to indications for continued PPI use were not adequately delineated or documented, and in some cases where there was not a clear symptomatic, endoscopic, or histologic benefit, PPIs were continued. In the two tertiary care referral centres with data analysed in this study, the most common reasons for continuation of PPIs both after a failed PPI trial and at the last documented patient contact were symptom improvement and, more surprisingly, lack of patient instruction to discontinue the medication. In long-term follow-up, >50% of EoE patients remained on PPIs after histologic non-response. Predictors of remaining on PPIs in the UNC cohort included male patients, patients with baseline heartburn symptoms, and patients with a hiatal hernia, while those with dysphagia were less likely to remain on PPIs. It is interesting to note that while most of the patients in this study were male, female subjects have been shown in prior studies to be more likely to discontinue medication or request discontinuation of medication^{15, 16} and thus results may be skewed by our male-predominant population and cannot be generalized to both male and female patients.

Data demonstrating high symptomatic and endoscopic response rates to PPIs in patients with EoE, despite histologic failure, may explain their prolonged use. In a large prospective study of response in EoE patients after 8 weeks of high-dose PPI therapy, PPIs induced clinical remission in >70% and endoscopic remission in >60% of patients, while histologic remission occurred in >50% of patients.¹⁷ This was borne out in a meta-analysis by Lucendo and colleagues which shows a higher symptom response rate than histologic response rate with PPIs,⁹ though it must be noted that symptom assessment in most studies of PPI use has not used validated patient-reported outcome measures. In our study, improvement in symptoms is a clinically appropriate justification for long-term PPI continuation.

There are several potential reasons for these higher symptomatic response rates. First, given that the prevalence of GERD in the general population is over 20%,¹⁸ EoE may occur with superimposed GERD.^{18, 19} Additionally, because of fibrostenosis and possible contractile dysfunction in EoE, there could also be secondary reflux or decreased clearance of physiologic reflux present,⁵ and therefore treating GERD would improve symptoms. Second, and related, there is significant overlap in symptoms of EoE and GERD. In a retrospective, case-controlled study evaluating the differences between EoE and GERD, the proportion of patients who reported heartburn was high in both the EoE and GERD groups (46% and 56%, respectively).²⁰ Similarly, in a prospective study evaluating prevalence of EoE in an adult population undergoing EGD, symptoms of chest pain, regurgitation, and nocturnal symptoms, symptoms often associated with GERD, were present in 40% or more patients with EoE,²¹ and related data have shown that these reflux-like symptoms are even more common in children.²² Again, if some of these symptoms are acid-mediated, then PPI treatment would provide symptom relief. In our study, symptom of heartburn and the presence of a hiatal hernia, both known to be correlated with GERD,^{20, 23} were significantly associated with remaining on PPI even after the diagnostic EGD for EoE. Similarly, in the UNC cohort, the persistence of dysphagia despite PPI use was significantly associated with PPI discontinuation. Third, pathologic reflux disease, associated with dilation of intracellular spaces and impairment in mucosal integrity,²⁴ may precipitate an inflammatory response leading to increased release of pro-inflammatory cytokines, recruitment of inflammatory cells to the oesophagus, and differentiation of TH2 cells to become allergen-specific.^{19, 25–27} Decreasing acid reflux with PPIs diminishes this chain of events.

In our study, we found that PPIs are often continued without a clear indication, and multiple systems issues may contribute. After diagnostic endoscopy, there is a lag in obtaining histology results that delay discussion and evaluation for continuing PPIs. At clinic follow-up, practitioners are operating under shorter clinic visits and demands of mandatory electronic charting.²⁸ For both of these reasons, PPI use may not be discussed. Furthermore, prior work has shown that general practitioners often do not feel empowered to change medications prescribed by a specialist and are thus less likely to decrease or discontinue them even when the indication isn’t clear, and patients may be hesitant to accept changes to their medications by a primary care provider.^{29, 30} Additionally, even if PPIs are discontinued appropriately, their accessibility as an over-the-counter medication could contribute to overuse.

Despite these barriers, it is essential to re-evaluate whether a clinical benefit to ongoing PPI use exists. While recent randomized, controlled data show few significant adverse effects of PPIs compared to placebo after 3 years and don’t justify discontinuation when appropriate indications exist,³¹ possible safety issues, when conveyed to a lay population, might drive patients to ask about stopping PPIs. It would be an interesting future direction to repeat our study in several years to see if the trends are different. Additionally, inappropriate PPI use increases the cost of EoE, which is already reported to be as high as hospital-related costs of acute appendicitis or GI bleeding, approaching or exceeding $1 billion annually.³² A recent multicentre study by Laserna-Mendieta et al examined efficacy of PPI in maintaining clinico-histologic remission in patients with EoE and found that 30% of patients experience loss of response. While loss of response with PPI is not as high as that seen with topical steroids³³, this finding reiterates the need for continued assessment of clinical benefit, particularly for patients on maintenance therapy.³⁴ Future studies might also examine how remission obtained on topical steroids versus dietary therapy impacts the chance of patients remaining on PPI therapy.

According to consensus guidelines, patients with EoE who have responded to and are maintained on PPIs should undergo routine evaluation for sustained benefit, particularly given the risk of waning effect as is seen with other EoE therapies.^{33, 35} Objective evidence and continued assessment of histologic, symptomatic, and endoscopic improvement should factor into decision-making for continuing PPI, and routine clinic visits are critical for monitoring for loss of clinical response.^{2, 14, 33, 36} Strategies during evaluation should incorporate knowledge of response rates of conventional therapy compared to PPIs, consideration of possible concomitant GERD, and the acknowledgement that the interplay between GERD and EoE is not yet fully understood. Future studies delineating EoE phenotypes may also help guide use of PPI in subsequent iterations of EoE guidelines.

There are several limitations of this study. As a retrospective study, we relied on documented reasons in the medical record for ongoing PPI use and could not prospectively assess the reasons by either the prescribing doctor or patient for continuing PPI. Lack of long-term follow-up with endoscopic re-evaluation may not capture the patients that over time developed a histologic response to PPI. As electronic medical record use was for clinical care, some charts had missing or ambiguous data and misclassification could be possible. While the presumption for this study was that providers always document their recommendations with clinic visits, the lack of documentation of PPI discontinuation does not necessarily imply therapy wasn’t discussed or stopped. Since the data are from tertiary care referral centres, the EoE patients and practice patterns may not be representative of all EoE care settings. However, these limitations are countered by strengths that include the two-centre study design, comprehensive and structured data extraction with standardized and coded chart abstraction, and a large patient sample.

In conclusion, this retrospective cohort study showed that a large proportion of EoE patients are continued on PPI therapy even after histologic non-response. While some of these patients had a clinical indication related to a combination of symptomatic, endoscopic, or (partial) histologic benefit, we found that many were often unintentionally kept on chronic PPIs for unclear reasons. Given that PPIs are now viewed as a treatment for EoE rather than a diagnostic criterion in the most recent consensus guidelines, our findings emphasize the need for consistent follow-up and re-evaluation when using PPIs for EoE treatment. In the context of questions about long-term PPI safety, a strong indication for continued use is particularly important.