Bytzer P., Jones R., Vakil N., Junghard O., Lind T., Wernersson B., Dent J. Limited ability of the proton-pump inhibitor test to identify patients with gastroesophageal reflux disease. Clin Gastroenterol Hepatol 2012; 10: 1360–1366.
Limited ability of the proton-pump inhibitor test to identify patients with gastroesophageal reflux disease
Peter Bytzer1, Roger Jones2, Nimish Vakil3, Ola Junghard4,†, Tore Lind4, Börje Wernersson4, and John Dent51 Department of Medicine, Køge University Hospital, Køge, Denmark;
2 Department of General Practice and Primary Care, King’s College London, London, United Kingdom;
3 Department of Medicine, University of Wisconsin Medical School, Madison, Wisconsin;
4 AstraZeneca Research and Development, Mölndal, Sweden; and
5 Department of Gastroenterology and Hepatology, Royal Adelaide Hospital and University of Adelaide, Adelaide, South Australia, Australia
Background & Aims: The efficacy of proton-pump inhibitor (PPI) therapy often is assessed to determine whether patients’ symptoms are acid-related and if patients have gastroesophageal reflux disease (GERD), although the accuracy of this approach is questionable. We evaluated the diagnostic performance of the PPI test, in conjunction with other tests, for the diagnosis of GERD.
Methods: We analyzed data from the DIAMOND study, a multinational trial that compared the ability of the reflux disease questionnaire with that of symptom-based clinical diagnosis to identify GERD in primary care patients with frequent upper-gastrointestinal symptoms. Patients (n = 308) were given placebo and further evaluated by endoscopy, wireless esophageal pH-metry, and symptom association monitoring. Those with GERD (n =197) were identified based on the presence of reflux esophagitis, esophageal pH level less than 4 for more than 5.5% of 24 hours, or positive results from symptom association monitoring (or a positive result from the PPI test in patients with borderline levels of esophageal acidity). All patients then were given single-blind therapy with esomeprazole (40 mg once daily) for 2 weeks and symptoms were recorded daily.
Results: A positive response to the PPI test was observed in 69% of patients with GERD and in 51% of those without GERD. Response to placebo did not influence the diagnostic ability of the subsequent PPI test. More patients with reflux esophagitis had a positive result from the PPI test than patients without GERD (57% vs 35%; P = .002) or patients with GERD but no esophagitis. A clinical diagnosis by the primary care physician of an acid-related disease was not associated with response to PPIs.
Conclusions: In a well-characterized population of primary care patients with frequent upper-gastrointestinal symptoms of any type, the PPI test has limited ability to identify patients with GERD, diagnosed by current standard tests. (ClinicalTrials.gov Number, NCT00291746.)
Keywords: Dyspepsia; Acid Reflux; Esophagitis; Esomeprazole.
Abbreviations used in this paper: GERD, gastroesophageal reflux disease; GI, gastrointestinal; LR, likelihood ratio; PCP, primary care physicians; PPI, proton-pump inhibitor; RDQ, Reflux Disease Questionnaire; SAP, symptom association monitoring.
© 2012 by the AGA Institute
In routine clinical practice, empiric acid suppression in the form of the proton-pump inhibitor (PPI) test often is used to determine whether upper-gastrointestinal (GI) symptoms are acid-related and whether gastroesophageal reflux disease (GERD) is present.1 However, its accuracy as a diagnostic tool has been questioned in some studies.2,3 The PPI test was used in the large international DIAMOND study in conjunction with other tests and symptom analysis to determine whether GERD was present in patients consulting in primary care with upper-GI symptoms.4 The DIAMOND study was unique in its study of a large population of primary care patients with frequent upper-GI symptoms. The main aim was to assess the accuracy of the Reflux Disease Questionnaire5 (RDQ) as a diagnostic tool for GERD and to compare it with endoscopy, pH-metry, and, in cases of borderline high esophageal acid exposure, also a positive PPI test. The overall results of the DIAMOND study showed that the RDQ, primary care physicians (PCP), and gastroenterologists have moderate and similar accuracy for diagnosis of GERD in a primary care population with frequent upper-GI symptoms. It also showed that a 2-week course of acid suppression did not add to the diagnostic precision when based on the predefined analyses. In this study, we examined various definitions of the response to PPI and the responses obtained in various patient subgroups and compared this with investigation-based criteria for GERD. Our aim was to investigate the diagnostic performance of the PPI test for prediction of GERD in this large population of patients reporting in primary care with frequent upper-GI symptoms.
MethodsWe analyzed data from the DIAMOND study, a multinational trial (Northwestern Europe, Canada) that tested the ability of the RDQ in diagnosing GERD among patients consulting for frequent upper-GI symptoms in primary care. Detailed information on recruitment, study design, and main results have been published elsewhere.4
In brief, a total of 73 PCP clinics screened 706 patients, who consulted with bothersome upper-GI symptoms. A total of 507 patients were enrolled; of these, 308 patients were fully evaluable. All patients completed the RDQ at entry and the investigators were blinded to this response. The PCP and the study gastroenterologist acting independently selected a symptombased diagnosis from a prespecified list. After this, the patient was asked to identify the most and second-most bothersome upper-GI symptom from a predetermined list of 19 symptoms. During treatment with placebo, identical in appearance to esomeprazole (patients were blinded as to when placebo or esomeprazole were administered), all patients had endoscopy with wireless 48-hour esophageal pH and symptom association monitoring (SAP). After investigation, patients received active esomeprazole 40 mg daily for 2 weeks (the PPI test) and the effect of treatment on their most and second-most bothersome symptoms (any type including symptoms typical for GERD, eg, heartburn) were recorded in daily diaries. Based on the outcome of investigation, all patients were classified as GERD or non- GERD according to predefined criteria. GERD was diagnosed when at least one of the following 4 criteria was present: (1) reflux esophagitis (any Los Angeles grade), (2) esophageal pH level less than 4 for more than 5.5% of the time, (3) positive SAP ( ≥95%) for association of symptoms with acid reflux, and (4) borderline high esophageal acid exposure (pH < 4 for 3.5%–5.5% of the time) and positive response of reflux-related symptoms to esomeprazole treatment (PPI test criteria)
A total of 203 patients were diagnosed with GERD using the investigation-based criteria. In 9 of these 203 patients, a diagnosis of GERD was based exclusively on a positive PPI test in patients with borderline high esophageal acid exposure. For the original publication,4 we performed exploratory analyses to examine the influence of excluding the PPI test criterion from the diagnosis of GERD and found only minimal change in the diagnostic yield of GERD. The possible influence of selecting different end points for a positive/negative PPI test on diagnostic yield also was explored because the DIAMOND study allowed patients to select their most bothersome and secondmost bothersome upper-GI symptoms, regardless of whether they were typical for GERD or not. Absence of the most bothersome symptom in the last 3 days of active treatment (esomeprazole 40 mg daily for 2 weeks) was a priori defined to be a positive response to the PPI test. In our exploratory analyses, we examined different combinations of symptoms typical for GERD (heartburn, central chest pain, dysphagia, or regurgitation) as the most or second-most bothersome symptoms and also restricted the analysis to the most bothersome symptom, to evaluate strategies for assessing whether a PPI test is positive or negative for reflux disease.
The placebo period was introduced to allow the endoscopy and pH-metry to take place and to form a run-in period for the PPI test to quantify the proportion of placebo responders. We analyzed outcome of the PPI test in placebo responders (absence of most bothersome symptom during the last 2 days of the placebo period) and nonresponders.
The likelihood ratio (LR) indicates the usefulness of a test in making a diagnosis. Values greater than 1 are associated with increasing likelihood of the disease.6 We analyzed positive (LR+) and negative (LR—) values for all patients and for subgroups.
All authors had full access to the study database.
Prevalence of Investigation-Based Gastroesophageal Reflux Disease and Response to Proton Pump Inhibitor Based on Heartburn, Central Chest Pain, or Regurgitation as the Most or Second-Most Bothersome Symptoms
The outcome of the PPI test was available for 299 of the 308 patients. As illustrated in Table 1, excluding the positive PPI test in borderline pathologic pH criterion for GERD did not change the diagnostic precision, with 66% (197 of 299) of patients being diagnosed with GERD on investigation, compared with 63% (188 of 299) when the positive PPI test in borderline pathologic pH criterion was excluded from the diagnosis. This is not surprising because only 9 patients were defined as having GERD on the basis of this criterion alone, and we therefore decided not to classify these patients as GERD in most of our post hoc analyses.
More patients with GERD had a symptom typical for GERD (eg, heartburn, regurgitation) as their most or second-most bothersome symptom compared with patients without GERD on investigation (69% vs 38%; Table 1). Dysphagia was not included as a symptom typical for GERD in these analyses because of the low numbers of patients reporting this as their most or second-most bothersome symptom.
When the definition of symptoms used for the PPI test was restricted to heartburn or central chest pain as the most or second-most bothersome symptoms (Table 1), the proportions of patients with a positive PPI test were similar to those obtained when regurgitation also was included in the assessment (Table 1). As expected, the number of non-GERD PPI responders increased slightly when the PPI test criterion was removed from the diagnostic criteria (Table 1).
Table 1. Positive PPI Test Results in Patients With Typical Reflux Symptoms Diagnosed With GERD According to the Protocol, or by Excluding the PPI Test From the Diagnostic Criteria
NOTE. A positive test was defined as the absence of the symptom during the last 3 days on PPI.
a Defined as a positive PPI test in patients with borderline abnormal esophageal acid exposure (pH < 4 for 3.5%–5.5% of the 24 hours).
Figure 1 summarizes the response of the most bothersome symptom in GERD and non-GERD patients, regardless of whether it is considered typical or atypical for GERD.
Response to Proton Pump Inhibitor Based on Heartburn, Central Chest Pain, or Regurgitation as the Most Bothersome Symptom
When the analysis took into account only the most bothersome symptom (any of heartburn, central chest pain, or regurgitation) to define the PPI test response, the proportion of patients with a positive PPI test was similar for patients diagnosed with GERD (on investigation) whether or not the positive PPI test in patients with borderline pathologic pH was included as a diagnostic criterion and irrespective of symptom (Table 1). The data from Table 1 indicate that the addition of a positive PPI test in patients with borderline pathologic pH to the GERD diagnostic criteria made very little difference to the final GERD diagnosis.
Response to Placebo
Patients were on placebo for a median of 8 days (interquartile range, 6–9 d) before they were switched to active esomeprazole. Response to placebo was defined as absence of the most bothersome and/or second-most bothersome symptom during the last 2 days of placebo treatment. This was analyzed for both GERD and non-GERD patients (Table 2). The proportions of patients in these groups who were symptom- free for the most bothersome symptom were 13% and 14%, respectively, with no important differences when analyzed for individual symptoms (Table 2). A similar analysis for the second- most bothersome symptom showed response rates overall of 17% in both GERD and non-GERD patients.
Table 2. Response to Placebo Defined as Complete Relief of Most Bothersome Symptom During the Last 2 Days on Placebo
Time to Response
Patients were treated with esomeprazole for a median of 13 days (IQ range, 13–15 d). We analyzed the time to complete symptom resolution in GERD and non-GERD patients. The proportion of patients with symptom resolution day by day is depicted in Figure 2. In both GERD and non-GERD patients the symptom response (heartburn, chest pain, regurgitation) to acid suppression increased during the first 5 to 6 days, with only minor fluctuations beyond that time point.
Response to Proton Pump Inhibitor in Gastroesophageal Reflux Disease and Non–Gastroesophageal Reflux Disease Patients Based on Endoscopic Classification
In 296 patients the response to the PPI test was analyzed according to the presence or absence of reflux esophagitis. We also analyzed the PPI test outcome in placebo responders and nonresponders and by most bothersome symptom. Significantly more patients with reflux esophagitis had a positive response to PPI than did non-GERD patients (57% vs 35%; P = .002). This difference was found both in placebo nonresponders (54% vs 27%; P < 0.1;GI symptom) and in patients with a bothersome symptom typical for GERD (72% vs 56%). Furthermore, we found a slightly higher response in GERD patients with reflux esophagitis than in those without (Table 3).
Are Physicians Able to Predict Response to Proton Pump Inhibitor?
During the screening visit the primary care physician recorded a most likely symptom-based diagnosis for each patient from a protocol-specified list of diagnoses. We hypothesized that a clinical diagnosis of an acid-related disease, such as GERD or peptic ulcer disease, would reflect the anticipated response to PPI. We analyzed the PPI response in patients predicted by the PCP to have GERD compared with all other patients. A positive PPI test was seen in 51% (83 of 162) of patients with a clinical diagnosis of GERD, compared with 44% (59 of 134) of patients who were predicted to have non-GERD diseases.
An analysis that combined the PCP clinical diagnoses of GERD and peptic ulcer disease did not improve the prediction of PPI response. A total of 48% (85 of 178) of patients predicted by the PCP to have GERD or peptic ulcer disease responded to PPI compared with 47% (55 of 118) of patients predicted to have other diseases (eg, dyspepsia).
Does the Proton Pump Inhibitor Test Provide Useful Support for the Gastroesophageal Reflux Disease Diagnosis?
The diagnostic characteristics of the PPI test were examined in terms of its sensitivity and specificity and positive and negative predictive value, in all patients, in placebo nonresponders, and in patients with symptoms typical of GERD. The LRs for a positive predictive PPI test (LR+) were 1.52, 1.81, and 1.26 for the 3 groups, respectively. LRs for a negative predictive PPI test (LR-) were 0.71, 0.69, and 0.67, respectively, for the 3 groups. The variation in LR between +2 and -2 for all subgroups suggested a very limited diagnostic value of the PPI test.
We previously showed that the probability of GERD increases with increasing scores on the RDQ.4 We hypothesized that a positive or negative response to PPI might help diagnose or rule out GERD in patients with ambiguous RDQ scores. Table 4 shows the positive PPI response for the most bothersome symptom grouped according to RDQ scores and symptom characteristic at entry. Both in patients in whom a symptom typical for GERD (eg, heartburn) was the most bothersome and in patients with a most bothersome symptom that was atypical, the probability of GERD increased with increasing RDQ scores at entry. However, response to PPI did not differ according to RDQ scores but was affected by the type of baseline symptom with higher PPI response rates seen in patients with a symptom typical for GERD as their most bothersome symptom (eg, heartburn) (66%–69%) compared with patients with another type of upper-GI symptom (eg, epigastric pain) (29%–38%). Thus, the probability of GERD in a patient with a most bothersome symptom not typical for GERD is related directly to the RDQ score at entry but not to the outcome of a PPI test. The RDQ was used among other tools to evaluate GERD symptoms in the study and we analyzed the changes in symptom dimensions recorded after PPI treatment in groups diagnosed on investigation with GERD or non-GERD. Heartburn and regurgitation showed a significantly greater improvement after PPI therapy in patients diagnosed with GERD, compared with non-GERD (change in respective mean RDQ scores, 1.44 vs 1.00, P < .02 for heartburn; and 1.56 vs 1.16, P = .02 for regurgitation), whereas changes in epigastric pain scores did not differ between the 2 groups (1.58 vs 1.57; P = .95).
DiscussionThe high prevalence of GERD makes it a costly health care issue for which the use of diagnostic tools such as endoscopy frequently is questioned for the diagnostic benefits it gives in relation to cost. The PPI test has been widely used clinically to assess whether upper-GI symptoms are acid-related and, as such, it has been promoted as a useful guide to treatment in patients with typical GERD symptoms and other acid-related diseases, without the need for endoscopy.7,8 However, prior studies have found that the PPI test has quite a high sensitivity for GERD but a poor specificity, and its value as a diagnostic test is limited when compared with pH monitoring and SAP analysis.9 The DIAMOND study was designed to overcome limitations in prior studies of the diagnostic value of a PPI test. Compared with previous trials, the DIAMOND study addressed the issue in a relevant study population of patients attending primary care for any upper-GI symptom and was not enriched with reflux disease patients. Furthermore, our study was characterized by state-of-the-art testing for the presence of reflux disease, daily symptom assessments, and well-defined definitions of response. This large database provides the best platform so far to test whether there are better approaches for scoring a test of PPI therapy. In these exploratory analyses we found that in a well-characterized population of primary care patients with frequent upper-GI symptoms of any type, the PPI test has limited ability to correctly predict GERD diagnosed by state-of-the art tests.
Similarly poor diagnostic ability was found when the PPI test was assessed in general practice alongside endoscopic findings and pH monitoring.10 In a meta-analysis of 15 published studies, Numans et al1 examined the sensitivity and specificity of the PPI test in relation to other objective measures of GERD such as pH monitoring and upper-GI endoscopy. The definition of a symptomatic response to a PPI was based on the criteria used in the individual studies. In general, these definitions were selected on the basis that they represented success in clinical practice and usually this involved “complete relief of heartburn.” Although up to 90% of the patients suspected of having uncomplicated GERD did respond to the empiric PPI test, their favorable response was not necessarily in agreement with the diagnosis established according to the traditional objective criteria. In addition, some of those diagnosed with GERD objectively did not show a good response to PPI. We acknowledge that the weakness of the association between proven GERD and a positive PPI test also may reflect the weaknesses of current standards for diagnosing GERD by investigation. A similar view was expressed by Monés11 in a review considering the diagnostic value of potent acid inhibition in GERD. On the other hand, the cost of treating all patients, some of whom will have a placebo response to any therapy, has not been evaluated adequately.
The DIAMOND study has provided data with a number of significant strengths. It is well powered and its results are generalizable to other primary care populations with a low prevalence of Helicobacter pylori infection. Furthermore, the study was based on several different and independent assessments of symptoms and included the best possible diagnostic approaches for GERD. By exploring different symptom groupings for the PPI test results, we had hoped to improve the diagnostic yield of this test. As found in the original DIAMOND study analyses, however, the results of our additional exploratory analyses showed a disappointingly low validity of the PPI test for the diagnosis of GERD as attested by the low positive and negative LRs in all patients and in various subgroups. In line with other researchers,1,12 we found that more patients with reflux esophagitis had a positive response to PPI than patients with GERD without esophagitis (57% vs 49%; Table 3). However, the clinical value of this is very limited because it is not possible to reliably separate reflux esophagitis from nonerosive reflux disease based on symptoms alone. Furthermore, a Norwegian study in 52 patients with heartburn or regurgitation as their main symptom found that a 7-day PPI test was unable to predict the outcome of a 24-hour pH-metry in patients with a normal endoscopy.13
It is an important strength of our study that it was performed in sequential primary care patients consulting for any upper-GI symptom. Only patients who had not used PPIs within 2 months before their first visit were included and all patients underwent comprehensive diagnostic testing to separate patients into GERD and non-GERD diagnoses. Previous trials on the diagnostic value of a short PPI test have evaluated the test against endoscopy alone14–17 or pH-metry alone,7,9,18–21 or only included selected patients referred for secondary/tertiary care evaluation.7,16
Can PCPs predict who will respond to a PPI? We hypothesized that a clinical diagnosis of GERD or peptic ulcer disease would reflect a higher likelihood of a positive PPI test. At least one study has shown a larger therapeutic response (PPI response rate minus placebo response rate) in patients entered into a clinical trial based on the primary care physician’s clinical judgment that the patients’ complaints were acid-related.22 However, a clinical diagnosis of GERD or peptic ulcer disease was not associated with a higher likelihood of a positive PPI response in the present study. A total of 48% of patients predicted by the primary care physician to have GERD or peptic ulcer disease responded to PPI compared with 47% of patients predicted to have other diseases. Furthermore, the use of the PPI test in conjunction with a diagnostic questionnaire (RDQ) proved to be of little value. Response to PPI did not differ according to baseline RDQ scores but was affected by symptom type, with higher response rates in patients with symptoms typical for GERD (eg, heartburn) compared with patients with symptoms atypical for GERD and therefore more likely to be related to other disorders.
There were some potential limitations to our study. A diagnosis of GERD was based on the outcome of the PPI test in some patients who had a borderline pH-metry. However, only 9 patients were classified as GERD based on this PPI test response criterion alone and all analyses were unaffected by the removal or reclassification of these 9 patients. We evaluated the outcome after 2 weeks of PPI, which has been the usual design in previous PPI test studies. Although we cannot rule out that a longer treatment period or a higher PPI dose would have provided better diagnostic values, previous studies and our own analysis shown in Figure 1 suggest that the sensitivity of the test stabilizes after the first 5 days of treatment.23 The use of complete symptom relief as an outcome measure in the PPI test was chosen primarily because it is scientifically attractive and not biased by the methodologic difficulties in measuring subtle changes over time in symptom severity, frequency, and duration. A less rigorous end point, such as symptom improvement, may be clinically more relevant and would have increased sensitivity, but at the expense of a higher placebo response. Finally, the inclusion of a placebo run-in period may be criticized as being artificial compared with everyday clinical practice. However, subgroup analyses in placebo responders and nonresponders did not show any impact on the test characteristics. All currently available diagnostic tests for GERD have limitations. Despite the use of multiple tests for the diagnosis of GERD, it is possible that some patients with GERD may have been misclassified and could have had a response to the PPI test. Reflux can be episodic and a 48-hour pH study may miss some patients who have reflux disease. Similarly, it is well known that many patients with reflux disease have a normal endoscopy.
In conclusion, the PPI test is likely to be positive in the majority of patients with typical reflux symptoms, but is not a definitive test for GERD because a significant proportion of non-GERD patients also have a positive PPI test. Even when combined with clinical diagnosis or a diagnostic questionnaire, the outcome of a PPI test does not add reliable information to distinguish between patients with and without GERD. Thus, in a well-characterized population of primary care patients with frequent upper-GI symptoms of any type, the PPI test has limited ability to predict GERD correctly.
Назад в раздел
Популярно о болезнях ЖКТ читайте в разделе "Пациентам"
Информация на сайте www.GastroScan.ru предназначена для образовательных и научных целей. Условия использования.