Accuracy of measurements of the length of esophageal lesions: An experimental study

• Introduction
• Methods
• The artificial esophagus
• Experimental design
• Statistical analysis
• Results
• Discussion
• References

Background and study aims:Hiatal hernia, Barrett’s esophagus and erosive esophagitis are defined and classified by measuring their length during endoscopy. The primary aim of our study was to evaluate the accuracy of length measurement of esophageal lesions with the conventional gastroscope marked at 5-cm intervals, and the secondary aim was to test the performance of a modified gastroscope marked at 1-cm intervals.

Methods: 24 senior endoscopists carried out endoscopy on a plastic mannequin. The esophagus was randomly replaced by one of a set of seven plastic tubes, each tube having two colored rings which were 1, 1.5, 2, 2.5, 3, 3.5 or 4 cm apart. The endoscopists were asked to measure the distance between the mouth-guard and each of the two ”esophageal” rings during endoscope withdrawal, with a precision of 0.5 cm. All participants endoscoped all seven tubes blindly, first using the conventional scope and then the modified scope.

Results: Using the conventional gastroscope, measurements were overall incorrect in 67.9 % (95 % CI 61 - 75.1) of cases and incorrect by at least 1 cm in 21 % (95 % CI 14.8 - 27.2) of cases. These percentages were significantly reduced by using the modified gastroscope (47.6 %, 95 % CI 40.1 - 55.2 and 7.1 %, 95 % CI 3.2 - 11.0, respectively) (P < 0.001). Overall accuracies in measuring lengths of ≥ 2 cm and ≥ 3 cm, corresponding to hiatal hernia and Barrett’s esophagus definitions, were 83 % and 94 % - 95 % with the conventional and modified gastroscopes, respectively (P < 0.05).

Conclusions: Our data suggest that estimation of the length of esophageal lesions using conventional endoscopes is inaccurate. The accuracy of measurements is significantly improved with the use of an endoscope marked at intervals of 1 cm.

Introduction

Upper gastrointestinal endoscopy is the main tool for diagnosis and classification of esophageal lesions. The length of the lesion is used for definition of hiatal hernia, classification of erosive esophagitis, and diagnosis of short- or long-segment Barrett’s esophagus. Indeed, hiatal hernia is defined as the protrusion of 2 cm or more, of the esophagogastric junction above the diaphragmatic impression [1]. The classification of erosive esophagitis according to the Los Angeles system is based on the length of mucosal lesions, known as mucosal breaks [2]. Barrett’s esophagus is recognised by the displacement of the squamocolumnar junction proximal to the esophagogastric junction and is classified as long- or short-segment if the above distance is more or less than 3 cm, respectively. [3]. The recently proposed ”C and M” classification of Barrett’s esophagus [4] requires even finer and more precise endoscopic measurement. Moreover, the length of the metaplastic segment may be associated with the risk of adenocarcinoma [5].

The measurement of the length of esophageal lesions is done using the numerical markings on the shaft of the endoscope. More precisely, with the endoscope tip located at the distal end of the lesion, the endoscopist reads the numerical mark on the endoscope shaft at the incisors, then withdraws the scope so that the tip is at the proximal margin of the lesion and then reads the new value at the incisors. The subtraction of the two readings gives the length of the lesion.

Our experience and published data [6] [7] suggest that these measurements could be inaccurate in a substantial proportion of cases. Therefore, the primary aim of our experimental ex vivo study was to investigate the accuracy of these measurements using a conventional endoscope, and the secondary aim was to examine whether the use of a modified endoscope with shaft markings at 1-cm intervals would improve accuracy.

Methods

The artificial esophagus

Endoscopies were done in a plastic mannequin (Keymed, Southend-on-Sea, Essex, UK) designed for training beginners in endoscopy. The mannequin was placed in the left lateral position and had the neck flexed and an adapted mouth-guard. Both the mannequin and the mouth-guard were held in place by an assistant. The ”esophagus” of this mannequin was randomly replaced, in turn, by seven different plastic tubes of 2.5 cm diameter. Each tube had two coloured rings separated by distances of 1, 1.5, 2, 2.5, 3, 3.5, or 4 cm. Each plastic tube was fitted into the mannequin esophagus so that the proximal esophageal ring of the tube varied in location between 37 and 41 cm from the mouth-guard ([Fig. 1]).

Experimental design

Endoscopy was carried out in the esophagus of the mannequin using a conventional gastroscope with 5-cm markings on the shaft (Olympus GIF-Q145) and with a modified instrument marked on the shaft at 1-cm intervals. For this a transparent membrane was wrapped tightly round the shaft of a conventional endoscope and circumferential lines were drawn on the membrane at 1-cm intervals ([Fig. 2]).

A total of 24 endoscopists, each with experience of more than 2000 gastroscopies participated in the experiments. Each endoscopist performed 14 esophagoscopies in two sessions, 1 month apart. In the first session they used the conventional gastroscope and in the second session the modified instrument. The endoscopists were blinded to the actual distances between the two ”esophageal” rings and were asked to measure the distance from the mouth-guard to each of the two esophageal rings during endoscope withdrawal, with a precision of 0.5 cm. The plastic tubes were randomly selected and fitted to the mannequin by one of the investigators who was unaware of the subsequent measurements. Measurements reported by the endoscopists were recorded by a different investigator. Endoscopists were not informed about their performance on the measurements until after the completion of the study. Data processing was performed by a third investigator who was blinded to the type of endoscope, i. e. conventional or modified, that had been used.

Statistical analysis

Multifactorial analysis of variance (ANOVA) was used to test whether the net error in measurements (real length minus measured length) varied among individual endoscopists.

Results are presented as percentages with 95 % confidence intervals (95 % CIs).

Significant differences in the percentage of incorrect measurements between the two methods (i. e. conventional vs. modified endoscope) were assessed using the Χ² test. Nonparametric statistics, i. e., Wilcoxon or Kruskall-Wallis tests, were used to compare differences in the median error of measurements (real length between the two rings minus measured length) between the two methods. Sensitivity, specificity, predictive values, and diagnostic accuracy were determined separately for the conventional and the modified gastroscope, for measuring lengths of ≥ 2 cm and ≥ 3 cm within the esophagus of the mannequin. The level of significance was defined as P < 0.05.

Results

Each one of the 24 endoscopists performed 14 esophagoscopies, seven with the conventional instrument and seven with the modified endoscope; thus in total there were 336 esophagoscopies, 168 with each type of endoscope.

There was not any significant variation among individual endoscopists regarding the net error in measurements (real length minus measured length), with either method (F ratio = 1.18, P = 0.26, with the conventional gastroscope; F ratio = 0.92, P = 0.6, with the modified gastroscope).

Using the conventional gastroscope, 67.9 % (95 % CI 60.95 - 75.05) of the measurements were incorrect, i. e., the estimated distance between the two rings was not identical to the real distance and 21 % (95 %CI 14.8 - 27.2) were incorrect by at least 1 cm. With the use of the modified gastroscope, the percentage of incorrect measurements was significantly lower at 47.6 % (95 % CI 40.1 - 55.2) (Χ² = 13.28, P < 0.001), and the percentage of measurements incorrect by more than 1 cm dropped to 7.1 % (95 % CI 3.2 - 11) (Χ² = 11.97, P = 0.002) ([Fig. 3]).

Overall, the median net error (real length between the two rings minus measured length) was significantly higher with the conventional gastroscope in comparison with the modified scope (Wilcoxon, W = 10 694, P < 0.005).

The median error in the measurement of the distance between the rings was significantly higher for the conventional instrument compared with the modified gastroscope for actual ring distances of 2 cm (T = 5.64, P = 0.018), 3 cm (T = 5.23, P = 0.022) and 4 cm (T = 4.11, P = 0.043), but not for any other ring distances ([Fig. 4]).

The performance of endoscopists in measuring lengths ≥ 2 cm (hypothetical cases of hiatal hernia) and ≥ 3 cm (hypothetical cases of long-segment Barrett’s esophagus) was significantly better when using the modified as compared with the conventional gastroscope ([Tables 1] and [2]).

Discussion

Our study addressed the issue of accuracy of length measurement of esophageal lesions during endoscopy. This is clinically important since the international definitions of Barrett’s esophagus, erosive esophagitis, and hiatal hernia rely on accurate endoscopic measurement. To our knowledge, this is the first study to assess the accuracy of length measurements during endoscopy objectively, based on known lengths in ex vivo experiments. Our data clearly show that 68 % of the measurements are incorrect and, more importantly, the error is 1 cm or more in one-fifth of the cases where a conventional gastroscope is used.

One may argue that the 0.5-cm precision that was required from endoscopists is an unrealistic goal, but it is relevant in clinical practice because 0.5 cm is the cutoff length of mucosal breaks for categorizing grade A or B esophagitis according to the Los Angeles classification system. However, our data have clearly shown that a precision of 1 cm is realistic when using the modified endoscope marked at 1-cm intervals, since only 7 % of measurements were erroneous.

Our data also show that the unsatisfactory performance was not related to interindividual variation, but was obviously related to the instrument properties, that is, the 5-cm intervals between the marks, which make length measurement difficult. This is because the endoscopist relies on reading off the marks on the shaft while the tip of the instrument is at the distal and proximal ends of the lesion and, since there are marks only every 5 cm, the endoscopist must make an estimate using the closest visible marking outside the model’s mouth. The unsatisfactory performance of the conventional endoscope observed in our experiments in the mannequin implies that in vivo endoscopic evaluation of the length of esophageal lesions is even more unreliable, as the conditions in real-life endoscopy are much more unfavorable. The mobility of the esophageal anatomical landmarks with respiratory movements or gagging, the variable flexion of the patient’s neck, the fact that the instrument may not be fully straightened and the occasional difficulty in the recognition of the landmarks, especially the proximal end of the gastric folds, are all factors that impede precise length measurements.

There are few published studies that have investigated the reliability of endoscopic measurement of the length of esophageal lesions in vivo. Kim et al. reported that, at a 6-week repeat endoscopy, 10 % of patients with Barrett’s esophagus showed a change in the relevant landmarks that was greater than 4 cm [6]. Dekel et al. found that the mean difference in length of Barrett’s esophagus between two consecutive endoscopies was 1.6 cm for patients who were followed up by the same endoscopist and 1.4 cm for patients followed up by different endoscopists. In addition, the longer the Barrett’s segment, the bigger the disagreement between the two consecutive measurements. For every 1-cm increase in the mean length of Barrett’s esophagus, the expected increase in the difference between two consecutive measurements was 0.15 cm [7].

The modified endoscope marked at 1-cm intervals had a significantly better overall accuracy. In our study, its use yielded measurements that were incorrect by more than 1 cm in only 7.1 % of cases. Eisen et al. used this type of endoscope together with toluidine blue staining and 1-cm interval biopsies for mapping metaplastic Barrett’s epithelium; this method had good reproducibility at follow-up testing, but the study was flawed because of the lack of a formal control group [8]. Earlier, in an attempt to obtain accurate real-time length measurement, Wakabayashi et al. proposed the use of a sophisticated computerized system including an electric catheter that was attached to the endoscope, which yielded very high accuracy [9]. However, we were not able to find any subsequently published evaluation of this system in humans.

In our study we have also investigated the performance of endoscopists using the conventional gastroscope and the modified instrument in diagnosing hypothetical hiatal hernia (esophageal ring distance of 2 cm or more) or hypothetical long-segment Barrett’s esophagus (esophageal ring distance of 3 cm or more) ([Tables 1] and [2]). In both situations, the endoscopists’ overall diagnostic accuracy was significantly better with the modified gastroscope compared with the conventional scope, with misclassification of only 6 % of hiatal hernias and 5 % of Barrett esophagus cases.

In conclusion, our data have shown that measuring the length of esophageal lesions using an endoscope with conventional 5-cm interval marking is a crude method even under ideal conditions of endoscopy. It is prone to notable error, of greater than 1 cm in one fifth of cases, and this may be of clinical significance as esophageal lesions such as hiatal hernia, Barrett’s esophagus, and erosive esophagitis can be misclassified. The use of the modified endoscope with 1-cm interval marking significantly improves the accuracy of these measurements by endoscopists, and the construction of such an instrument by manufacturers would be only minimally more expensive than that of a conventional endoscope.

Competing interests: None

References

• 1 Hill L D, Kozarek R A, Kraemer S J. et al . The gastroesophageal flap valve: in vitro and in vivo observations.  Gastrointest Endosc. 1996;  44 541-547
  CrossrefPubMedSearch in Google Scholar
• 2 Armstrong D, Bennett J R, Blum AL. et al . The endoscopic assessment of esophagitis: a progress report on observer agreement.  Gastroenterology. 1996;  111 85-92
  CrossrefPubMedSearch in Google Scholar
• 3 Sharma P, McQaid K, Dent J. et al . AGA Chicago Workshop. A critical review of the diagnosis and management of Barrett’s esophagus: the AGA Chicago Workshop.  Gastroenterology. 2004;  127 310-330
  CrossrefPubMedSearch in Google Scholar
• 4 Sharma P, Dent J, Armstrong D. et al . The development and validation of an endoscopic grading system for Barrett’s esophagus: The Prague C & M criteria.  Gastroenterology. 2006;  131 1392-1399
  CrossrefPubMedSearch in Google Scholar
• 5 Avidan B, Sonnenberg A, Schnell T. et al . Hiatal hernia size, Barrett’s length, and severity of acid reflux are all risk factors for esophageal adenocarcinoma.  Am J Gastroenterol. 2002;  97 1930-1936
  CrossrefPubMedSearch in Google Scholar
• 6 Kim S L, Waring J P, Spechler S J. et al . Diagnostic inconsistencies in Barrett’s esophagus.  Gastroenterology. 1994;  107 945-949
  PubMedSearch in Google Scholar
• 7 Dekel R, Wakelin D E, Wendel C. et al . Progression or regression of Barrett’s esophagus - is it all in the eye of the beholder?.  Am J Gastroenterol. 2003;  98 2612-2615
  PubMedSearch in Google Scholar
• 8 Eisen G M, Montgomery E A, Azumi N. et al . Qualitative mapping of Barrett’s metaplasia: a prerequisite for intervention trials.  Gastrointest Endosc. 1999;  50 814-818
  CrossrefPubMedSearch in Google Scholar
• 9 Wakabayashi T, Nakazawa S, Yoshino J. et al . A new method of real-time endoscopic measurement with an electric catheter.  . 1994;  26 466-469
  PubMedSearch in Google Scholar

S. D. Ladas, MD

Hepato-Gastroenterology Unit

Attikon University General Hospital

1 Rimini Street

12462 AthensGreece

Fax: + 30-210-5326422

Email: sdladas@hol.gr

References

• 1 Hill L D, Kozarek R A, Kraemer S J. et al . The gastroesophageal flap valve: in vitro and in vivo observations.  Gastrointest Endosc. 1996;  44 541-547
  CrossrefPubMedSearch in Google Scholar
• 2 Armstrong D, Bennett J R, Blum AL. et al . The endoscopic assessment of esophagitis: a progress report on observer agreement.  Gastroenterology. 1996;  111 85-92
  CrossrefPubMedSearch in Google Scholar
• 3 Sharma P, McQaid K, Dent J. et al . AGA Chicago Workshop. A critical review of the diagnosis and management of Barrett’s esophagus: the AGA Chicago Workshop.  Gastroenterology. 2004;  127 310-330
  CrossrefPubMedSearch in Google Scholar
• 4 Sharma P, Dent J, Armstrong D. et al . The development and validation of an endoscopic grading system for Barrett’s esophagus: The Prague C & M criteria.  Gastroenterology. 2006;  131 1392-1399
  CrossrefPubMedSearch in Google Scholar
• 5 Avidan B, Sonnenberg A, Schnell T. et al . Hiatal hernia size, Barrett’s length, and severity of acid reflux are all risk factors for esophageal adenocarcinoma.  Am J Gastroenterol. 2002;  97 1930-1936
  CrossrefPubMedSearch in Google Scholar
• 6 Kim S L, Waring J P, Spechler S J. et al . Diagnostic inconsistencies in Barrett’s esophagus.  Gastroenterology. 1994;  107 945-949
  PubMedSearch in Google Scholar
• 7 Dekel R, Wakelin D E, Wendel C. et al . Progression or regression of Barrett’s esophagus - is it all in the eye of the beholder?.  Am J Gastroenterol. 2003;  98 2612-2615
  PubMedSearch in Google Scholar
• 8 Eisen G M, Montgomery E A, Azumi N. et al . Qualitative mapping of Barrett’s metaplasia: a prerequisite for intervention trials.  Gastrointest Endosc. 1999;  50 814-818
  CrossrefPubMedSearch in Google Scholar
• 9 Wakabayashi T, Nakazawa S, Yoshino J. et al . A new method of real-time endoscopic measurement with an electric catheter.  . 1994;  26 466-469
  PubMedSearch in Google Scholar

S. D. Ladas, MD

Hepato-Gastroenterology Unit

Attikon University General Hospital

1 Rimini Street

12462 AthensGreece

Fax: + 30-210-5326422

Email: sdladas@hol.gr