Magnification endoscopy for diagnosis of nonerosive reflux disease: a proposal of diagnostic criteria and critical analysis of observer variability

• Introduction
• Patients and methods
• Study design
• Patients
• Symptom questionnaire
• Esophageal pH monitoring
• Endoscopic procedure
• Image processing
• Evaluation process for proposed criteria
• Image handling
• Statistical analysis
• Results
• Characteristics of the study subjects and patients
• Internet assessment
• Direct assessment
• Discussion
• Acknowledgments
• References

Background and study aims: This study tested the diagnostic value of high-resolution endoscopy for the recognition of subtle diagnostic esophageal mucosal changes in nonerosive reflux disease.

Patients and methods: Ten control subjects and eleven patients with nonerosive reflux disease confirmed by a validated questionnaire, standard endoscopy, and 24-hour pH-metry participated in the study. Still images were collected by high-resolution endoscopes from the distal esophagus in a standardized manner, incorporating iodine staining. Assessments were repeated in the patients with reflux disease after 4 weeks of esomeprazole therapy. Interobserver variability in the recognition of the proposed criteria was initially evaluated by 27 endoscopists using an Internet-based process. After optimisation of image quality the evaluation was repeated face-to-face with six expert endoscopists.

Results: No criterion was identified in either assessment that was sufficiently sensitive and specific to patients with reflux disease to be clinically useful. The kappa value, used to assess interobserver variation, was acceptably high only for invisibility of palisade vessels (0.59). Triangular indentations, apical mucosal breaks, and pinpoint blood vessels at the squamocolumnar junction were identified more frequently in the patients with reflux disease (P < 0.05). These changes and the invisibility of the palisade vessels were significantly less prevalent in reflux patients after therapy (P < 0.01).

Conclusions: Though some distal esophageal mucosal appearances observed with the high-resolution endoscope appeared to be related to nonerosive esophageal mucosal injury, none of these changes proved to be sufficiently sensitive and specific to justify their use as a diagnostic criterion for nonerosive reflux disease.

Introduction

The utility of endoscopy for the assessement of severity of reflux disease has been increased by the development of the Los Angeles classification of reflux esophagitis by the International Working Group for the Classification of Esophagitis [1,2]. For the significant proportion of patients with reflux disease who do not have mucosal breaks, endoscopy would only be diagnostically helpful if reflux disease caused endoscopic features other than mucosal breaks that could be reliably recognized. There is still a widespread belief that esophageal mucosal changes such as patchy erythema, friability, edema and blurring of the squamocolumnar junction are useful criteria for the endoscopic diagnosis of reflux disease, but formal testing has shown that endoscopists cannot agree on the presence or absence of such minimal changes [1,3].

Endoscopic criteria are warranted for the diagnosis of gastroesophageal reflux disease (GERD) in cases that are now classified as “endoscopy-negative” or “nonerosive” reflux disease (NERD). This study was prompted by the introduction of a new generation of high-resolution endoscopes with magnification capabilities that make it possible to see the mucosa in greater detail than was possible previously, so with the potential for reliable recognition of nonerosive peptic damage of the esophageal squamous mucosa and consequent improvement in the sensitivity of endoscopy for the diagnosis of GERD [4]. Accordingly, the present study was undertaken with high-resolution magnifying endoscopes with the following aims: (a) identification of potential endoscopic criteria for recognition of nonerosive acid-peptic squamous mucosal injury; (b) evaluation of the prevalence of similar changes in a carefully defined control group of subjects without reflux disease and in patients with proved nonerosive reflux disease; (c) determination of whether these changes resolved after 4 weeks of very effective acid suppression; and (d) assessment of interobserver agreement for recognition of these proposed criteria by expert endoscopists.

Patients and methods

Study design

We recruited healthy subjects and untreated patients who had significant reflux symptoms, but no esophageal mucosal breaks observed at conventional endoscopy. High-resolution endoscopy was carried out according to a standard protocol in both groups, including the structured gathering of high-quality still images at defined positions around the gastroesophageal junction. Digital storage of these images meant that they could be reproduced without any loss of quality. The healthy volunteers were not evaluated further, but the patients with reflux disease were then treated with the proton-pump inhibitor (PPI), esomeprazole 40 mg each morning for 4 weeks, after which another high-resolution endoscopy was performed, using the same protocol as for the first endoscopy.

The best representative still images for each high-resolution endoscopy examination were then selected. Each of these individual patient image sets was then coded and stored in an electronic file. Once the proposed criteria had been developed, the image sets were sorted into random order and blinded before evaluation by expert endoscopists.

Patients

Prior to entry into the study, in both the reflux disease patients (n = 11) and in the healthy subjects (n = 10), upper gastrointestinal symptoms were evaluated using a self-completed questionnaire [5]. To be eligible for the study, the healthy subjects were required to be free of any gastrointestinal symptoms. To be accepted as a reflux disease patient, an abnormal reflux symptom score in the questionnaire was required (see below). Whenever possible, esophageal pH monitoring was performed to enhance the validity of the classification of asymptomatic subjects as free of reflux disease and to strengthen the diagnosis of NERD.

The study was approved by the human research ethical committees at the University of Gothenburg, Academisch Medisch Centrum in Amsterdam, and the Royal Adelaide Hospital.

Symptom questionnaire

The symptom questionnaire used in this study was originally developed for the symptom-based diagnosis of reflux disease [5]. Validated translations were available in all three languages required for this study (Swedish, Dutch, English). A score of 4 or more was regarded as positive for reflux disease [5].

Esophageal pH monitoring

A monocrystalline antimony pH electrode was positioned 5 cm above the lower esophageal sphincter as determined by manometry [6]. Recordings were stored in a data logger and each dataset was analyzed using standard commercial software (Medtronic). An esophageal pH of less than 4 during at least 4.0 % of the measured time or during at least 5 % of the time when the subject was in the upright position was regarded as pathological.

Endoscopic procedure

The initial upper endoscopy was performed with a standard-resolution endoscope (GIF-100; Olympus, Tokyo, Japan), and after this a high-resolution magnification endoscope (EG485-ZH; Fujinon Co., Omiya, Japan) was used to evaluate the esophagus and gastroesophageal junction further. This endoscope has a resolution of 850 000 pixels with an optical magnification of × 35 and a × 2 electronic zoom. Hyoscine butylbromide (20 - 40 mg) and midazolam (1 - 2 mg) was given intravenously according to individual needs. The distal esophageal mucosa was thoroughly investigated with and without optical magnification.

Digital, uncompressed, high-resolution images were captured in each subject as follows. Firstly, images were taken without magnification to obtain an overview of the region. Images were then captured under magnification, 1 - 2 cm proximal to the squamocolumnar junction, as well as in at least two of the four quadrants (at the 3-, 6-, 9- and 12-o’clock positions) at the squamocolumnar junction. After this, the esophageal mucosa was flushed with 10 - 20 mL of acetylcysteine (20 mg/mL) and then rinsed with water. Lugol’s iodine solution (2 %, 10 - 20 mL) was sprayed over the distal esophageal mucosal surface (using an Olympus PW5L spray catheter) and after rinsing with water, iodine-stained images were collected according to the same protocol as described above [7].

Image processing

All images were captured and stored initially by the Fujinon EF101 software program, which stored images in uncompressed JPEG format. These images were subsequently converted into BITMAP format and transferred over the Internet to a shared database on a server in Gothenburg, Sweden. Four to six of the best-quality images per high-resolution endoscopy procedure from each of the 29 endoscopies were selected for subsequent evaluation by the assessors. The selected images were required to give a good overview of the squamocolumnar junction as well as magnified views of at least two quadrants of the circumference of the gastroesophageal junction. At least one of the two visualized quadrants was required to be a good-quality image of iodine stained mucosa. Furthermore an unstained magnified image 1 - 2 cm proximal to the squamocolumnar junction was required. All image sets were randomly distributed and coded. The code was not broken until all analyses had been completed.

Evaluation process for proposed criteria

The nine proposed high resolution endoscopy criteria for nonerosive squamous mucosal injury by gastroesophageal reflux were derived from a pilot study [8]. These criteria were first presented and explained to 11 endoscopists who were members of the International Working Group for the Classification of Esophagitis, after which the coded images were displayed to the expert group and the consistency and agreement between assessors were explored. As a result of this pilot exercise, the following criteria for reflux disease were judged to be potentially useful and so were chosen for evaluation in the present study: triangular indentations into the squamous mucosa as an extension from villiform columnar mucosa at the squamocolumnar junction ([Figure 1 a]); minute apical mucosal breaks at the vertex of a triangular indentation ([Figure 1 a]); the visibility of palisade blood vessels, which were defined as longitudinal blood vessels seen in squamous mucosa just above the squamocolumnar junction ([Figure 1 b]); (when this was not the case, this was considered a possible marker of superficial mucosal opacification due to edema); the presence of pinpoint or comma-shaped blood vessels seen in squamous mucosa above the squamocolumnar junction ([Figure 1 c]); branching blood vessels seen in columnar mucosa just below the squamocolumnar junction ([Figure 1 d]); a serrated squamocolumnar junction, which was defined as more than three saw-tooth incursions into the squamous mucosa (with the depth of each saw tooth being greater or equal to its width) per radial gastric fold ([Figure 1 e]); and the presence of villiform mucosa, defined as villous-like mucosa immediately below the squamocolumnar junction ([Figure 1 f]).

Image handling

The image bank generated for scoring of the proposed criteria was evaluated in two separate exercises. In the first (the Internet assessment), 35 experienced endoscopists from seven countries were recruited to score the images. Training on the proposed criteria was provided for these endoscopists in the form of an explanatory Powerpoint presentation delivered either on a CD-ROM or via the Internet. This presentation included examples of the proposed criteria as itemized above. The image sets for assessment were delivered to the endoscopists via the Internet, with a score sheet. Endoscopists were instructed to make a yes/no judgment on the presence of each of the proposed criteria, within each image set.

The second evaluation (the direct assessment) was undertaken because of two major uncertainties about the effectiveness of the Internet assessment. Firstly it was thought possible that the reduction in the amount of information in the images that was required for their transmission over the Internet could have obscured changes that might have been recognizable in the original version of the images. Secondly, there was concern that the training of the 35 endoscopists for recognition of the proposed criteria had not been adequate. Accordingly, the direct assessment was done by gathering six endoscopists for a separate meeting at which an initial face-to-face explanation of the proposed criteria was provided, together with examples contained in specimen images. These specimen images and the image sets evaluated subsequently were full digital copies of the original images and were projected onto a screen for evaluation with equipment that preserved all of the information contained within the 21 images considered to be of the highest quality.

Statistical analysis

Levels of agreement were evaluated and expressed with kappa statistics, which take account of the possibility of agreement by chance [9]. All statistical analyses were performed using SAS statistical software. Kappa values of less than 0.20 signify poor agreement, values of 0.20 - 0.40 fair agreement, values of 0.41 - 0.60 moderate agreement, values of 0.61 - 0.80 good agreement, and values of 0.81 - 1.0 very good agreement. A kappa value of 1.0 indicates perfect agreement [9]. Additional analyses were performed by application of the Wilcoxon signed rank test for dependent variables and the Mann-Whitney test for independent variables.

Results

Characteristics of the study subjects and patients

Eleven patients with reflux disease (six men, five women, mean age 48) and ten healthy subjects (four men, six women, mean age 43) gave informed consent and were enrolled in the study. Esophageal pH monitoring was carried out successfully in 8/10 asymptomatic subjects and in all the patients with reflux disease. Data from individual subjects are summarized in [Table 1]. In six of the reflux patients, the total esophageal acid exposure time was ≥ 4 % of the recorded 24 hours; in one patient, acid exposure was within normal limits (3.5 % of the total time), but frequent reflux episodes were recorded (132 episodes per 24 hours), and the symptom association probability was positive (97.2 %) [10]; another reflux disease patient had a total 24-hour esophageal acid exposure time within the normal range, but a pathological upright reflux time (5 %), with a symptom association probability value of 95 %; esophageal pH monitoring was negative in the remaining three patients. It was not possible to perform 24-hour pH monitoring in two of the control subjects. The mean questionnaire score in the reflux group was 10.45 ([Table 1]). A second high-resolution endoscopy was completed in 8/11 patients; the other three patients declined the follow-up endoscopic procedure.

Internet assessment

Data from eight assessors were excluded from the analysis because these individuals scored fewer than 27 of the 29 image sets. The remaining 27 assessors contributed 351 observations for each of the proposed criteria. [Table 2] presents the data for the comparison of the prevalence of each criterion in the image sets from control subjects and for the reflux disease patients prior to therapy. In the patients, the highest prevalence for any finding was 65 %. All of the proposed changes were also judged to be present in the healthy control subjects at rates that differed relatively little from the reflux disease patients. No criterion was present significantly more frequently in the untreated reflux disease patients than in the healthy controls.

[Table 2] also shows the data for the subsets of the six asymptomatic subjects in whom pH monitoring confirmed the absence of high levels of acid reflux and for the eight reflux disease patients in whom pathological acid exposure was confirmed (see above and [Table 1]). In this subanalysis, there was a numerically improved separation for the presence of several criteria in the reflux disease patients, but the only criterion that showed a statistically significant difference in level of occurrence when compared with the control subjects was the presence of triangular lesions at the gastroesophageal junction (P < 0.05). The prevalences of the proposed criteria in the reflux disease patients before and after esomeprazole therapy are also given in [Table 2]. These data are presented for all eight patients for whom such data existed and for the six patients in whom esophageal pH monitoring was unequivocally positive. A paired analysis revealed that the prevalences of four criteria changed significantly after esomeprazole therapy. While triangular indentations and apical mucosal breaks became a less frequent finding, branching blood vessels below the squamocolumnar junction became more frequent. These differences became even more marked after consideration of the levels of acid exposure, when the absence of palisade blood vessels above the squamocolumnar junction was also found to be significantly less prevalent after esomeprazole therapy.

At the initial calculation of the percentage of agreement between the expert endoscopists, all tested criteria ranged from 54 % to 77 %, with apical mucosal breaks, pinpoint blood vessels, and branching blood vessels all over 70 % ([Table 2]). Interobserver variability, as assessed by the respective kappa values, displayed poor agreement or lack of agreement. In fact, only palisade and branching blood vessels reached kappa values above 0.2 ([Table 2]).

Direct assessment

[Table 3] presents the outcomes of the evaluations of the 21 selected image sets by the six endoscopists who participated in this second evaluation. In the case of the comparison between the primary groups of healthy subjects and untreated reflux disease patients, the findings differed little from those of the Internet assessment. In contrast, the data from the pH monitoring-defined subgroups indicated that triangular lesions, apical mucosal breaks, and the presence of pinpoint blood vessels in the squamous mucosa were present significantly more frequently in reflux disease patients than in healthy controls (P < 0.05). The sensitivity and specificity of these criteria, however, were low. The comparison of images taken in the reflux disease patients before and on completion of esomeprazole therapy gave the strongest support to the relevance of some of the criteria identified as possibly due to acid-peptic damage. For both the primary patient group and the subset of patients in whom esophageal pH monitoring had confirmed the symptom-based diagnosis of reflux disease, the prevalences of triangular lesions, apical mucosal breaks, the presence of pinpoint blood vessels in the squamous mucosa, and also the absence of palisade blood vessels in the squamous mucosa, were all significantly reduced following treatment with esomeprazole (P < 0.01 to P < 0.001) (see [Table 3]).

By and large, we also saw an improvement in the respective figures in the interobserver variability analysis. The percentage of agreement now converged towards a range between 60 % and 80 %. Regarding kappa values, the overall picture looked very much the same, signifying poor to fair agreement between the endoscopists, except for palisade blood vessels, which reached a kappa value of 0.59 ([Table 3]). We were unable to detect any important difference in outcomes when the reflux patients were selected solely on the basis of symptoms or on abnormal ambulatory 24-hour pH values.

Discussion

This study has proposed and investigated seven novel criteria that can be assessed using high-resolution magnifying endoscopy that could potentially improve the recognition of reflux disease. The major outcome is that it has failed to identify any new criterion that is of any practical value for this purpose.

At first sight, the differences in the frequencies with which some of the criteria were recognized in the patients with NERD compared with the control subjects appear to hold some promise. The kappa values for interobserver agreement on the presence or absence of the criteria tested were, however, unacceptably low, and it is this measure that will determine the diagnostic value of any test. It is evident therefore that all the criteria would have major limitations in terms of sensitivity and/or specificity.

The first step in the present study was the proposal of novel criteria that seemed to hold promise for testing. We based our choice of criteria on quite extensive preliminary observations with high-resolution endoscopes and on insights into the histological nature of nonerosive reflux-induced injury to the distal esophageal mucosa. It remains possible that there are changes that we have failed to identify that could be scored reliably at endoscopy and which would be useful criteria for recognition of nonerosive reflux-induced injury.

We took considerable trouble to rule reflux disease out in our healthy, asymptomatic control subjects and to firmly establish its presence in the patients by a structured symptom assessment, and by pH monitoring. Although pH monitoring is not a standard reference test, it is relatively sensitive for the presence or absence of reflux disease. Accordingly, we did a subgroup analysis confined to the asymptomatic control subjects who had a normal pH-monitoring test and to all the patients in whom pH monitoring was positive. Though this subgroup analysis was slightly more positive for the value of the criteria, the kappa values for recognition of different criteria derived from this subgroup assessment were not substantially different from the unsatisfactory values obtained in the entire population enrolled in the study. The criterion that did reach a moderate level of inter-individual agreement was the “palisade vessels not visible” criterion. Although this criterion did not show any difference in prevalence between patients with NERD and controls in the distal part of the esophagus, a significant difference in the prevalence was noted between the pre-PPI therapy and post-PPI therapy results. The addition of a placebo-controlled arm in our study would have strengthened the methodology and might have been able to answer the question of whether the appearance of visible palisade vessels was due to effective acid suppression.

The methods designed for gathering, distributing, and assessing endoscopic images represented significant technical challenges for a study such as this. Though we made strenuous efforts to optimize our approaches, we cannot exclude the possibility that these could have impacted negatively on the outcomes of assessments of the criteria. Concern that the delivery of the test images via the Internet could have influenced the results through loss of image quality led us to conduct a second evaluation using an approach that aimed to overcome this potential limitation. The best approach that is now technically possible for studies such as this would probably be to make high-quality digital video recordings of the entire examination of the esophagus for each endoscopy, using a mucosal visualization protocol similar to the one we employed in the present study. This would more closely emulate the “real time” approach of an endoscopist looking carefully for evidence of nonerosive mucosal injury.

The use of Lugol’s iodine staining is an additional potential option for recognition of nonerosive injury that has been explored recently in a study by Yoshikawa et al. [11]. In this study, unstained mucosal streaks were found in only 1/38 (3 %) controls, compared with 19/39 (49 %) patients with NERD (P < 0.0001). Our study protocol was not designed to evaluate the prevalence of such streaks, but we found that iodine-staining defects were associated with triangular indentations and apical mucosal breaks at the squamocolumnar junction (as can be seen in [Figure 1 a]).

Kiesslich et al. [12] recently presented their experience with high-resolution endoscopy in nonerosive reflux disease. From a cohort of 99 patients referred for endoscopy, 39 were classified as having NERD and another 39 paients who did not have overt symptoms of GERD served as controls. These investigators also submitted their patients to 4 weeks of PPI therapy and then performed another endoscopy. An over-representation of punctate erythema proximal to the squamocolumnar junction was found in patients with NERD and this responded to PPI therapy. As in our own experience, irrespective of the criteria used, these researchers were confronted with sensitivity and specificity problems. It is quite difficult to make direct comparisons between these two studies because of the huge differences between their study design and performance, but it is clear that further research into the vasculature patterns of the distal squamous epithelium in GERD is warranted.

In summary, this study has proposed and explored seven different distal esophageal mucosal appearances that can be observed with the high-resolution endoscope. Some appeared to be related to nonerosive esophageal mucosal injury, but none of the changes proved to be sufficiently sensitive and specific to justify their use as a diagnostic criterion for nonerosive reflux disease.

Acknowledgments

We wish to thank Ola Junghard at AstraZeneca and Emma Johnsson at Statistiska Konsultgruppen for expert statistical support. All the endoscopy specialists in Adelaide, Amsterdam, Gothenburg, Magdeburg, and Nantes who participated in the evaluation process are gratefully acknowledged. We are indebted to AstraZeneca and Fujinon for financial, logistic, and technical support for the study.

Competing interests: The study was financed by an unconditional grant from AstraZeneca and Fujinon (endoscopes).

References

• 1 Armstrong D, Bennett J R, Blum A L. et al . The endoscopic assessment of esophagitis: a progress report on observer agreement.  Gastroenterology. 1996;  111 85-92
  CrossrefPubMedSearch in Google Scholar
• 2 Lundell L R, Dent J, Bennett J R. et al . Endoscopic assessment of oesophagitis: clinical and functional correlates and further validation of the Los Angeles classification.  Gut. 1999;  45 172-180
  CrossrefPubMedSearch in Google Scholar
• 3 Walshe P, Rowley H, Hone S. et al . Is reflux noted at diagnostic rigid oesophagoscopy clinically significant?.  J Laryngol Otol. 2001;  115 552-554
  PubMedSearch in Google Scholar
• 4 Bruno M J. Magnification endoscopy, high resolution endoscopy, and chromoscopy: towards a better optical diagnosis.  Gut. 2003;  52 (Suppl 4) iv7-11
  CrossrefPubMedSearch in Google Scholar
• 5 Carlsson R, Dent J, Bolling-Sternevald E. et al . The usefulness of a structured questionnaire in the assessment of symptomatic gastroesophageal reflux disease.  Scand J Gastroenterol. 1998;  33 1023-1029
  CrossrefPubMedSearch in Google Scholar
• 6 Ruth M, Hamelin B, Rohss K. et al . The effect of mosapride, a novel prokinetic, on acid reflux variables in patients with gastro-oesophageal reflux disease.  Aliment Pharmacol Ther. 1998;  12 35-40
  CrossrefPubMedSearch in Google Scholar
• 7 Misumi A, Kondou H, Murakami A. et al . Endoscopic diagnosis of reflux esophagitis by the dye-spraying method.  Endoscopy. 1989;  21 1-6
  Thieme ConnectPubMedSearch in Google Scholar
• 8 Tam W. Proposed new endoscopic criteria for minimal change reflux esophagitis based on magnification endoscopy (abstract 3439).  Gastrointest Endosc. 2001;  53 AB119
  PubMedSearch in Google Scholar
• 9 Altman D G. Practical statistics for medical research. London; Chapman and Hall 1991
  Search in Google Scholar
• 10 Weusten B L, Roelofs J M, Akkermans L M. et al . The symptom-association probability: an improved method for symptom analysis of 24-hour esophageal pH data.  Gastroenterology. 1994;  107 1741-1745
  PubMedSearch in Google Scholar
• 11 Yoshikawa I, Yamasaki M, Yamasaki T. et al . Lugol chromoendoscopy as a diagnostic tool in so-called endoscopy-negative GERD.  Gastrointest Endosc. 2005;  62 698-703
  CrossrefPubMedSearch in Google Scholar
• 12 Kiesslich R, Kanzler S, Vieth M. et al . Minimal change esophagitis: prospective comparison of endoscopic and histological markers between patients with non-erosive reflux disease and normal controls using magnifying endoscopy.  Dig Dis. 2004;  22 221-227
  CrossrefPubMedSearch in Google Scholar

A. Edebo, MD

Department of Gastro-Surgical Research

Sahlgrenska University Hospital

413 45 Göteborg, Sweden

Fax: + 46-31-411882

Email: anders.edebo@surgery.gu.se

References

• 1 Armstrong D, Bennett J R, Blum A L. et al . The endoscopic assessment of esophagitis: a progress report on observer agreement.  Gastroenterology. 1996;  111 85-92
  CrossrefPubMedSearch in Google Scholar
• 2 Lundell L R, Dent J, Bennett J R. et al . Endoscopic assessment of oesophagitis: clinical and functional correlates and further validation of the Los Angeles classification.  Gut. 1999;  45 172-180
  CrossrefPubMedSearch in Google Scholar
• 3 Walshe P, Rowley H, Hone S. et al . Is reflux noted at diagnostic rigid oesophagoscopy clinically significant?.  J Laryngol Otol. 2001;  115 552-554
  PubMedSearch in Google Scholar
• 4 Bruno M J. Magnification endoscopy, high resolution endoscopy, and chromoscopy: towards a better optical diagnosis.  Gut. 2003;  52 (Suppl 4) iv7-11
  CrossrefPubMedSearch in Google Scholar
• 5 Carlsson R, Dent J, Bolling-Sternevald E. et al . The usefulness of a structured questionnaire in the assessment of symptomatic gastroesophageal reflux disease.  Scand J Gastroenterol. 1998;  33 1023-1029
  CrossrefPubMedSearch in Google Scholar
• 6 Ruth M, Hamelin B, Rohss K. et al . The effect of mosapride, a novel prokinetic, on acid reflux variables in patients with gastro-oesophageal reflux disease.  Aliment Pharmacol Ther. 1998;  12 35-40
  CrossrefPubMedSearch in Google Scholar
• 7 Misumi A, Kondou H, Murakami A. et al . Endoscopic diagnosis of reflux esophagitis by the dye-spraying method.  Endoscopy. 1989;  21 1-6
  Thieme ConnectPubMedSearch in Google Scholar
• 8 Tam W. Proposed new endoscopic criteria for minimal change reflux esophagitis based on magnification endoscopy (abstract 3439).  Gastrointest Endosc. 2001;  53 AB119
  PubMedSearch in Google Scholar
• 9 Altman D G. Practical statistics for medical research. London; Chapman and Hall 1991
  Search in Google Scholar
• 10 Weusten B L, Roelofs J M, Akkermans L M. et al . The symptom-association probability: an improved method for symptom analysis of 24-hour esophageal pH data.  Gastroenterology. 1994;  107 1741-1745
  PubMedSearch in Google Scholar
• 11 Yoshikawa I, Yamasaki M, Yamasaki T. et al . Lugol chromoendoscopy as a diagnostic tool in so-called endoscopy-negative GERD.  Gastrointest Endosc. 2005;  62 698-703
  CrossrefPubMedSearch in Google Scholar
• 12 Kiesslich R, Kanzler S, Vieth M. et al . Minimal change esophagitis: prospective comparison of endoscopic and histological markers between patients with non-erosive reflux disease and normal controls using magnifying endoscopy.  Dig Dis. 2004;  22 221-227
  CrossrefPubMedSearch in Google Scholar

A. Edebo, MD

Department of Gastro-Surgical Research

Sahlgrenska University Hospital

413 45 Göteborg, Sweden

Fax: + 46-31-411882

Email: anders.edebo@surgery.gu.se