Article Text
Abstract
Objectives In most patients undergoing endoscopy for upper gastrointestinal (GI) symptoms in the Western world, no macroscopic abnormality or evidence of Helicobacter pylori infection is identified. Following this negative investigation, proton pump inhibitor (PPI) therapy is usually prescribed. The aim of this study was to assess the value of such treatment compared with placebo and to identify predictors of response.
Design Prospective parallel randomised study.
Setting Dyspepsia Research Clinic.
Participant 105 patients (49 men, median age 44 years, IQR 22) with normal endoscopy and H pylori negative with ongoing upper GI symptoms following 2-week run-in period.
Intervention Full demographic symptom severity and characteristics were assessed and 24 h oesophageal pH metry and oesophageal manometry were performed prior to randomisation to 2 weeks of treatment with lansoprazole 30 mg/day or placebo (2:1), with reassessment of symptom severity during the second week of treatment.
Primary outcome 50% reduction in Glasgow Dyspepsia Severity Score (GDSS).
Results According to intention to treat analysis, the response was 35.7% for the active group and 5.7% for the placebo group (p<0001). The only non-invasive independent predictor of response to PPI in multivariable analysis was the patient's body mass index (BMI) (p=0.003). The association of BMI with response to PPI was apparent across the full range of quartiles (p values for trend=0.01). BMI had a similar predictive value to either 24 h oesophageal pH metry or manometry. Predominant symptom and symptom subgroups were unhelpful in predicting the response to PPI. Including all pretreatment assessments, only BMI (p<0.05) and lower oesophageal sphincter pressure (p<0.05) were independent predictors of response.
Conclusion The response to PPI therapy is likely to be related to underlying acid reflux. The strong predictive value of BMI is probably due to its association with underlying reflux disease and the fact that it is a more objective and reproducible measure than symptom characteristics. It is recommended that BMI should be measured in patients with upper GI symptoms.
Trial Registration Number ISRCTN 32863375.
- Dyspepsia
- upper gastrointestinal
- non-erosive reflux disease
- proton pump inhibitor
- symptoms
- BMI
- non-ulcer dyspepsia
- obesity
- proton pump inhibition
Statistics from Altmetric.com
- Dyspepsia
- upper gastrointestinal
- non-erosive reflux disease
- proton pump inhibitor
- symptoms
- BMI
- non-ulcer dyspepsia
- obesity
- proton pump inhibition
Significance of this study
What is already known on this subject?
In most patients with upper GI symptoms in the Western world, endoscopy is normal and they subsequently receive proton pump inhibitor (PPI) therapy.
The response to PPI therapy in these patients has been studied in those preselected as having non-ulcer dyspepsia or non-erosive reflux disease but not in all-comers.
Symptoms are known to have limited value in predicting response to PPI therapy.
What are the new findings?
In patients with non preselected upper GI symptoms plus normal endoscopy who are H pylori negative, PPI therapy is superior to placebo.
BMI is a highly significant predictor of response and similar to oesophageal pH metry or manometry.
Symptoms are of little value in predicting response in this group of patients.
How might it impact on clinical practice in the foreseeable future?
BMI should be assessed routinely in patients with upper GI symptoms and used as a predictor of the response to PPI therapy.
Introduction
Upper gastrointestinal (GI) symptoms are common and their management challenging. If endoscopy reveals peptic ulcer disease or reflux oesophagitis, specific treatment can be prescribed. However, in the majority of patients with upper GI symptoms, endoscopy shows no macroscopic abnormality to explain the symptoms or guide treatment.1 Helicobacter pylori infection may be confirmed in some of these patients and eradication therapy occasionally helps.2
An increasingly common challenge is the management of upper GI symptoms in patients with normal endoscopy and no evidence of H pylori infection or in those whose symptoms persist after eradication of the infection. Such patients are usually prescribed proton pump inhibitor (PPI) therapy. Previous studies of the efficacy of PPI therapy in patients with normal endoscopy have usually included a mixture of H pylori positive and negative patients. The previous studies have also focused on patients with either non-ulcer dyspepsia (NUD) or non-erosive reflux disease (NERD), with these subgroups being examined in separate studies. The NUD studies have included patients with upper GI symptoms and excluding heartburn as the predominant symptom.3–8 The NERD studies have included patients with typical reflux symptoms and sometimes accompanied by abnormal 24 h oesophageal pH metry.9 10 In those with NERD, the benefit of PPI over placebo has been demonstrated with a therapeutic gain of approximately 30%.9 10 In patients with NUD, the benefit of PPI versus placebo has been less with a therapeutic gain of approximately 10%.3–8 11 These studies indicate that the character of the upper GI symptoms has some (albeit small) predictive value regarding the likely response to PPI therapy.
Previous studies of the symptomatic response to PPI in patients with normal endoscopy are not entirely applicable to real life as such patients do not always fit into the two symptom subgroups previously studied. For example, patients with retrosternal discomfort would be excluded from the NUD group but might also not be included in the NERD group if their symptoms were not typical of heartburn/reflux. In clinical practice, patients usually undergo endoscopy on the basis of upper GI symptoms without their preclassification.
The response to PPI therapy in patients with normal endoscopy and no H pylori infection is likely to be related to gastro-oesophageal acid reflux. Over the past few years there has been increasing evidence for a strong role for body mass index (BMI) in the aetiology of reflux disease.12–14 However, previous studies have not investigated the value of BMI in predicting the response to PPI therapy or how it compares with the predominant symptom or invasive tests such as oesophageal pH metry and manometry.
The primary aim of this study was to compare PPI with placebo in patients referred for endoscopic investigation of upper GI symptoms and with normal endoscopic findings and no evidence of H pylori. The secondary aim was to assess the value of symptoms and other non-invasive information as well as 24 h pH metry and oesophageal manometry in predicting the response to PPI in this group of subjects.
Methods
Patients
The patients recruited into the study had been referred by their primary care physician to our dyspepsia research clinic for endoscopic investigation of upper GI symptoms of >3 months duration. The clinic was established to facilitate research into the diagnosis and management of upper GI disease. Although based in the hospital, the research clinic operates in close collaboration with primary care practices, providing them with open access endoscopy for their patients at an early stage in their management. The research clinic does not accept patients with sinister symptoms such as dysphagia, weight loss, persistent vomiting or evidence of GI bleeding.
Patients whose endoscopy showed no evidence of peptic ulcer disease, erosive oesophagitis or hiatus hernia and who were H pylori negative on histological examination and rapid urease test of antral biopsies were invited to participate in the study. Patients were excluded if they were taking non-steroidal anti-inflammatory drugs or other medication known to cause upper GI symptoms.
Study design
This was a parallel design randomised trial. Patients fulfilling the entry criteria had an initial assessment (visit 1) and their symptoms were then monitored over a 2-week run-in period. Those with significant ongoing symptoms on review at 2 weeks (visit 2) underwent oesophageal manometry and 24 h oesophageal pH metry prior to randomisation to 2 weeks treatment with lansoprazole 30 mg/day or placebo. Their symptoms were reassessed 2 weeks later (visit 3). The study thus involved attending the dyspepsia research clinic on three separate occasions as detailed below and summarised in the flow chart shown in figure 1.
Visit 1
On their first visit, patients' demographic data were recorded including history of smoking and alcohol consumption, current and previous medication and past history of ulcer disease or H pylori eradication therapy. They were also asked if they experienced stress at work or at home and for each the response was categorised as none, some or a lot. Their height and weight were recorded and BMI calculated. Abdominal examination was performed and recorded as normal or tender and the site of any tenderness recorded.
The character of their GI symptoms was recorded in two ways: (1) by asking them for their predominant symptom which was categorised as retrosternal pain/discomfort/regurgitation, epigastric pain/discomfort, fullness/satiety, nausea/vomiting, or other; and (2) by asking them if they had any other GI symptoms and these were categorised according to their location as retrosternal, epigastric, left upper quadrant, right upper quadrant, generalised, other, none. They were asked whether they had symptomatic benefit from antacids or antisecretory medication and the response categorised as usually, sometimes or never.
They were then asked a series of questions to determine the extent to which their symptoms fitted with those typically associated with the following four conditions: peptic ulcer, gastro-oesophageal reflux, dysmotility disorder and irritable bowel syndrome (IBS). The specific questions are shown in box 1.
Questionnaire for assessment of symptom subgroups
Ulcer-like questions
Is the pain relieved by eating?
Does it wake you through the night?
Is the pain worse before meals or when you are hungry?
Is the pain periodic?
Total ulcer-like score
Reflux-like questions
Is the symptom worse if you bend forward?
Is it worse when you first go to bed at night?
Is it worse after a large heavy meal?
Do you get regurgitation of stomach contents?
Do you find food sticks or gives you discomfort after swallowing?
Total reflux-like score
Dysmotility-like questions
Are you troubled with upper abdominal distension/bloating?
Do you feel full up after finishing a meal?
Are you troubled with belching?
Do you have a feeling of nausea?
Do you avoid a wide range of foods?
Total dysmotility-like score
Irritable bowel syndrome (IBS)-like questions
Do you find that your discomfort is relieved by emptying your bowels?
Do you find that you have to go to the toilet to empty your bowels more frequently when the discomfort starts?
Do you find that you have looser stools at the onset of the pain/discomfort?
Have you noticed if you pass any mucus in your stools?
After you have been to the toilet, do you experience a feeling of incomplete evacuation?
Do you experience a feeling of lower abdominal bloating?
Are your stools often loose and watery (25% of the time)?
Total IBS-like score
Note: Each question scores 2 for usually, 1 for sometimes and 0 for never, and the scores are then added together to give the total score for each symptom subgroup.
Following this assessment, the patients were given Gaviscon (Reckitt Benckiser, Hall, UK) to take as required and told to avoid any other medication for their symptoms. They were provided with a symptom diary to complete over the 7 days prior to their next visit. This recorded the frequency, severity and duration of symptoms and Gaviscon usage and allows calculation of the modified Glasgow Dyspepsia Severity Score (GDSS).15 16 The GDSS is a global severity score which records overall symptom severity independent of the character or site(s) of the symptom or symptoms. Arrangements were made for them to reattend in 2 weeks.
Visit 2
On review at the research clinic, the GDSS over the preceding 7 days was assessed by patient interview and review of their diary card. Any subject whose symptom score over the preceding 7 days was <3 was excluded from entering the randomised study.
At this visit, those with significant continuing symptoms underwent oesophageal pull-through station perfusion manometry of the lower oesophageal sphincter (LOS). The manometer catheter was then removed and an antimony pH catheter was placed 5 cm above the upper border of the LOS for recording oesophageal acid exposure over the subsequent 24 h. The patient used the event marker to indicate meals, supine versus erect posture and symptom episodes.
The patients returned 24 h later to have the pH catheter removed. On the same visit they were given a 2-week course of either lansoprazole 30 mg/day or identical placebo. This was split 2/3 active. The medication was blinded to both patients and investigators. Participants were allocated to either active or placebo medications through block randomisation conducted by hospital pharmacy using a computer random number generator. The patients were instructed to take one capsule each morning 30 min before breakfast and to fill out a further symptomatic diary over the 7 days prior to the next visit 2 weeks later.
Visit 3
At the third visit the GDSS for the preceding 7-day period was assessed by examining the diary card and directly interviewing the patient. Compliance with the study medication was assessed by performing a tablet count.
Measurement of LOS parameters
Manometry was performed with the patients fasted and supine using a water-perfused manometry catheter with four ports arranged at the tip of the catheter circumferentially 90 degrees apart and four ports at 5 cm intervals proximal to this. Each channel was perfused with sterile water at a rate of 0.6 ml/min using a low-compliance pneumohydraulic capillary infusion pump (Ardorfer Medical Specialties, Greendale, Wisconsin, USA). Each channel was connected to a pressure transducer positioned at the level of the mid-axillary line. The equipment was calibrated prior to each test by taking a pressure tracing with the ports held at the level of the mid-axillary line and another reading with the ports directly above the mid-axillary line, that height representing a hydrostatic pressure equivalent to 50 mm Hg. The catheter was advanced via the anaesthetised nostril into the oesophagus and stomach. A slow station pull through was then made at 1 cm withdrawal increments.
Analysis of the manometry recording was performed using Polygram (Medtronic, Synectics, New Hampshire, USA). The length of the LOS was assessed by measuring its distal end as the first position with a 2 mm Hg increase over the intragastric baseline pressure and its proximal border as the position where the pressure dropped to end expiratory baseline pressure. The length of the LOS was further subdivided into an abdominal length and thoracic length based on the position of the respiratory inversion point. The maximum end-expiratory LOS pressure was calculated by identifying the station recording maximal pressures and then using three end-expiratory pressures from each of the four circumferential ports to produce a mean (of 12) end expiratory maximum LOS pressure. This LOS pressure was expressed relative to the intragastric pressure. The intragastric pressure was expressed relative to the baseline calibration pressure recorded at the level of the mid-axillary line.
pH Metry
The oesophageal pH recordings were analysed as the percentage time pH <4 for total, upright and supine periods. Symptom index was calculated for those subjects with acid exposure pH<4 for <4.2% of the total 24 h recording period.17
Statistical analyses
The primary outcome of the study was the percentage change in GDSS. A power calculation was made to detect a minimum difference in symptom change of 20% between the two groups taking α=0.05, power=0.95, case/control ratio 2/1 and 25% within-group SD. The required sample size was 65 cases and 35 controls.
All baseline demographic data are presented as median (IQR). The differences between variables were tested using non-parametric methods.
To assess predictors of response to PPI, all variables with a significant relationship to the response in univariable analysis were entered into a multivariable logistic regression model. A stepwise forward likelihood method was used in the multivariable analysis. All variables including age, BMI, LOS pressure, intragastric pressure in expiration, oesophageal exposure to pH <4 in total time and supine time were treated as continuous variables. The combined variable of symptom index >50% and oesophageal exposure to pH <4, total time >4.2% was used as a dichotomous factor with a simple contrast method. The stepwise criteria for entry and removal of factors in each step were 0.05 and 0.10, respectively.
Most analyses were conducted using statistical software SPSS Version 17.
Results
One hundred and fifty-two patients were initially assessed but 47 were excluded at visit 2 as their symptom severity was <3. Of the 105 who were randomised, 3 withdrew during the treatment period (2 placebo, 1 active), leaving 102 patients who completed the study. The study was started in December 1997 and completed in January 2000 according to the predefined timetable.
The predominant symptom of those randomised was retrosternal pain/discomfort/regurgitation in 43%, epigastric pain in 42%, nausea/vomiting in 9.5%, other symptoms in 2.9% and three patients did not respond clearly to this question. Previous treatment included PPI therapy in 59%, histamine H2 antagonists in 17% and antacid only in 8.6%. Three per cent were underweight (BMI <18.5 kg/m2), 39% were within normal range (BMI 18.5–24.9 kg/m2), 38% were overweight (BMI 25–29.9 kg/m2) and 20% were obese (BMI ≥30 kg/m2). The pretreatment characteristics of the placebo and active group were similar, as shown in table 1.
The primary outcome (reduction in symptom severity score during days 7–14 of treatment) was significantly different between the active and placebo groups. The proportion of patients experiencing ≥50% reduction in their GDSS was 35.7% for PPI and 5.7% for placebo (p<0.001, Fisher exact test for intention to treat). The difference was also significant if expressed as absolute change in symptom score (median±IR 3±5 vs 1±2; p=0.000). No adverse effects were reported by either participants or investigators.
Further analysis was performed to determine whether there were any significant predictors of response to PPI therapy in 66 patients treated with PPI. Univariable analysis indicated that increasing age, BMI, intragastric pressure, oesophageal acid exposure (both total and supine) and decreasing LOS pressure and dysmotility-like symptom score were predictors of a positive response to PPI as continuous variables (table 2). The univariable predictive value of 24 h oesophageal pH <4 for >4.2% of the time was not improved by combining it with symptom index >50% in those with <4.2% acid exposure. When expressed as dichotomous variables, BMI (<25 kg/m2≥), LOS pressure (<10 mm Hg≥) and oesophageal pH <4 (≤4.2%>) were also significant predictors in the univariable analysis.
The predominant symptom was unhelpful in predicting the response in the univariable analysis. Those whose predominant symptom was retrosternal pain/discomfort/regurgitation showed a similar response to active treatment compared with those with other predominant symptoms (p=0.21). Patient scores for each of the three symptom complexes (ie, reflux-like symptoms, ulcer-like symptoms and IBS-like symptoms) showed no association with response to PPI. As mentioned above, the dysmotility-like score was a weak negative predictor of response. Those with two or more secondary symptoms showed a lower response rate than those with fewer secondary symptoms, but there was no significant association between the number of secondary symptoms and the response to PPI when the former was assessed as a continuous variable.
When multivariable analyses were performed excluding the invasive investigations of oesophageal pH metry and manometry, the only significant predictor was BMI (p=0.003, R2=0.25, table 3). Applying the multivariable analysis model to both non-invasive and invasive parameters showed only BMI and LOS pressure as independent predictors of response in the final step.
Comparison was made of the predictive value of the only independent non-invasive predictor (BMI) with that of the two independent invasive predictors (oesophageal pH metry and manometry). To assess this the patients were divided dichotomously based on the median value for each of these parameters and the value of the higher versus lower group was compared with respect to the response to PPI. As seen in figure 2, BMI had a similar predictive response to oesophageal pH metry or manometry. The association between BMI and the response to PPI was apparent across the full range of BMI quartiles (p value for trend=0.011, figure 3).
Discussion
In North America and Western Europe, the majority of subjects undergoing endoscopy for upper GI symptoms have no macroscopic abnormality or evidence of H pylori infection. Our study confirms that PPI therapy is superior to placebo in such patients and that the therapeutic gain is substantial at 30%.
The mechanism of the response to acid suppression is likely to be largely or entirely related to the symptoms being due to acid reflux. Normal endoscopy and the absence of H pylori infection excludes any other established mechanism of acid-induced pain. Further support for this is the association of the response to PPI with abnormal oesophageal 24 h pH metry and/or LOS pressure on manometry.
Previous studies assessing the response to PPI therapy in subjects with upper GI symptoms and normal endoscopy have consisted of studies in patients with NERD9 10 (ie, with typical reflux symptoms) or separate studies in patients with NUD 3–8 11 (ie, with upper GI symptoms excluding predominant heartburn). The NUD studies have shown a therapeutic gain of approximately 10% and the NERD studies a therapeutic gain of the order of 30%.3–11
Our study, which included all patients referred for endoscopic investigation of uncomplicated upper GI symptoms, produced a greater therapeutic gain than might have been anticipated from a group comprised largely of a mixture of patients with NUD and NERD. This higher therapeutic gain may be explained by including only patients with significant ongoing symptoms following the 2-week run-in period and thus minimising the placebo response. We recognise that our study is from a single centre in the Western world and its generalisability may be limited to similar populations.
Symptom characteristics were of little value in predicting the response to PPI therapy. In particular, the therapeutic gain in those with predominantly retrosternal discomfort/pain/regurgitation was similar to that in those with other predominant symptoms. A high dysmotility-like score was a weak negative predictor of response in univariable analysis. None of the other details obtained in the history including stress at home or work was useful in predicting the response to PPI therapy. A substantially larger study might have identified symptom characteristics which showed a statistically significant association with response to PPI treatment, but the value of these predictors would have remained small.
Previous studies have assessed the ability of symptoms to predict a response to PPI in patients with upper GI symptoms and normal endoscopy. In a report by Bolling-Sternevald of 826 patients (from two separate studies) with epigastric pain/discomfort and normal endoscopy, positive predictors included age >40 years, bothersome heartburn, low bloating scores, epigastric pain and diarrhoea, <3-month history and low improvement of vitality.18 However, although these characteristics were statistically significant, the magnitude of their effect was very small and the authors concluded that the pretreatment symptoms would provide little help in clinical practice. Likewise, in a study by Peura et al of 921 patients with non-heartburn-predominant upper abdominal pain/discomfort, a superior response was reported in those with some heartburn but again the actual difference in response was small at 10%.6 In a study by Bytzer et al which included 217 patients with non-heartburn-predominant non-ulcer dyspepsia, none of the patients' symptoms or characteristics were predictive of a response to PPI therapy.19 In a further study by Bolling-Sternevald et al of 197 patients with non-heartburn-predominant upper GI pain/discomfort, the only symptom characteristic predicting a positive response was night-time pain, but again the magnitude of the effect was small and considered to be of little clinical significance.7 The poor predictive response of symptoms is consistent with their poor ability to predict the nature of upper GI disease.20
Carlsson et al attempted to produce a questionnaire which would more reliably identify reflux disease and predict response to PPI therapy.21 The description of ‘a burning feeling rising from the stomach or lower chest towards the neck’ was the only question which was helpful in identifying reflux disease. However, in their 187 patients with normal endoscopy and upper GI symptoms which could include predominant heartburn, a response to PPI was seen in 57% of those with the symptom and 33% without it, with a placebo response also of 33%. Even this symptom characteristic was therefore of relatively little help in predicting the response to PPI therapy.
In contrast to the lack of value of symptoms, we found that the patient's BMI was a highly significant independent predictor of response to PPI therapy. Indeed, it was the only non-invasive independent predictor of response. Patients in the upper half of BMI (≥25 kg/m2) were 3.9 times more likely to respond than those in the lower half. The link between BMI and the response to PPI was seen across the full range of BMI values. The magnitude of the predictive value of BMI was also greater than that previously reported for the predictive value of symptom characteristics in patients with normal endoscopy. The 52.9% response for those in the upper half of the BMI range versus 22.6% for those in the lower half of the BMI range compares favourably with even the best previously reported predictive value of the classical heartburn description by Carlsson et al showing a 59% response in those with the symptom versus 33% for those without it.21
In our study, the predictive value of BMI was not only superior to symptoms but similar to that from the invasive investigations of 24 h pH metry or oesophageal manometry. Interestingly, previous studies of predictors of the response to PPI have not assessed the value of the BMI.6 7 18 19 21 This may be explained by the fact that most of the studies were conducted before the important role of BMI in reflux disease was recognised. It will be important to assess the value of BMI in predicting the response to PPI therapy in other patient populations such as in the primary care setting.
It seems likely that the ability of BMI to predict the response to PPI therapy reflects its association with reflux disease and thus provides a surrogate marker of the likelihood of underlying acid reflux. A number of mechanisms are thought to explain the association between BMI and acid reflux. BMI increases intra-abdominal pressure and gastro-oesophageal pressure gradient and this is thought to predispose to hiatus hernia and to increased volume of refluxate during sphincter incompetence.22 23 The frequency of transient LOS relaxation and the proportion associated with acid reflux are also positively correlated with BMI.24 Patients with increased BMI have a higher prevalence of hiatus hernia and of defective LOS function.23 25 BMI might also be a surrogate marker for dietary habits which provoke acid reflux. The nature of BMI itself may contribute to its predictive value. In contrast to the highly subjective and variable nature of upper GI symptoms, BMI is objective and reproducible.
In conclusion, our study demonstrates the value of BMI in predicting the response to PPI in patients with normal endoscopic findings. In addition, it reconfirms the relative lack of value of the patient's symptom characteristics in predicting response. We recommend routinely assessing BMI in patients with upper GI symptoms as a useful non-invasive indicator of their likely underlying cause.
References
Footnotes
Funding This study received partial financial support from Wyeth Pharmaceuticals.
Competing interests None.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the Greater Glasgow Health Board – The West Ethical Committee.
Provenance and peer review Not commissioned; externally peer reviewed.