Manometry is a technique commonly used to evaluate. Value of Spatiotemporal Representation of Manometric Data. Methods Subjects

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2008;6: Value of Spatiotemporal Representation of Manometric Data CLAUDIA GRÜBEL,* RICHARD HISCOCK, and GEOFF HEBBARD* *Department of Gastroenterology, The Royal Melbourne Hospital, Parkville, Australia; and Department of Obstetrics and Gynaecology, Mercy Hospital for Women, Heidelberg, Australia Background & Aims: High-resolution manometry with spatiotemporal representation of pressure data is a technique that has developed during the past years. We compared spatiotemporal and traditional line plot representation of manometry data in a group of medical students in terms of the ability of the user to come to a rapid and accurate diagnosis and evaluated user preferences for the 2 forms of data display. Methods: After standardized paperbased and electronic tutorials, 60 medical students classified 30 typical examples of a range of motility disorders in both line plots (10 sensors, including a virtual sleeve ) and spatiotemporal plot format (derived from 16 sensors). Swallows were presented electronically in random order. The accuracy and speed of the assessment were compared between the 2 forms of data presentation, as well as a subjective rating of preference. Results are presented as mean standard error of the mean. Results: Classifications based on data presented in spatiotemporal format were more often correct (89% 1.2% vs 86% 1.3%, P.002), and correct diagnoses were provided more promptly ( seconds vs seconds, P <.001) than in line plot format. Sixty-eight percent of the study population preferred the spatiotemporal presentation. Conclusions: The analysis of manometry data by manometry-naïve individuals is faster and more accurate when data are presented in spatiotemporal than in line plot format. In addition, users preferred the spatiotemporal plots. Spatiotemporal presentation of manometric data is likely to be more easily understood by patients and the non-expert physician community. Manometry is a technique commonly used to evaluate gastrointestinal motility disorders. During the past 15 years, high-resolution manometric assemblies have been developed that allow detailed assessment of pressures along the length of the esophagus and LES. Initially used only for research studies or in tertiary centers, these have now become available for routine clinical use, and electronic data acquisition has largely replaced paper-based recording. The first use of computers in manometric data acquisition was to display data in a similar method to the chart recorders that they replaced, that is, as a series of line plots (LPs), with each plot representing one pressure sensor. With the development of high-resolution manometry (HRM), it was soon realized that the computer was capable of displaying the data in other formats that might offer advantages in terms of interpretation. One of these is to display data as a 3-dimensional map of pressure plotted against both time and position within the esophagus (topographic or spatiotemporal representation). 1 This method of data display is believed by those who use it to be an easily accessible and efficient form of data representation, which might lead to more rapid or accurate diagnosis or understanding of esophageal motility disorders; however, this has never been formally evaluated, 2 6 and it remains uncertain to what degree the benefits are due to the increased amount of pressure information present or the mode of presentation as a spatiotemporal plot (STP). A study in a group of experienced gastroenterologists comparing the accuracy of STPs and LPs has shown that HRM with spatiotemporal representation of the manometric data demonstrates features not easily appreciated on standard LPs and allowed diagnosis in patients who had previously proved challenging. 7 Nevertheless, experienced users of conventional manometry are often skeptical of the value of HRM and spatiotemporal representation of manometry data and do not believe that it is likely to improve their diagnostic accuracy. 8 In contrast, new users who have not been exposed to manometric data often comment that they find the spatiotemporal representation easier to understand once the principles are explained to them, and they find this representation more intuitive than the traditional representation as a series of LPs. Although an expert making an accurate diagnosis is clearly a central aspect of esophageal manometry, communication of the findings and understanding of esophageal physiology by colleagues and patients are also important in clinical practice. The aim of this study was to compare spatiotemporal data display with a standard LP representation in terms of the ability of a manometry-naïve user to come to a rapid and accurate diagnosis and to evaluate user preferences for the 2 forms of data display. We hypothesized that the presentation of data in the spatiotemporal format would allow subjects to make correct assessments in a greater proportion of cases, and that these would be made more quickly. Methods Subjects Sixty medical students were recruited from Melbourne and Monash Universities. Students had grounding in the basics of physiology and variable degrees of clinical exposure but no significant previous exposure to esophageal manometry. The study was approved by the Human Research Ethics Committee Abbreviations used in this paper: DES, diffuse esophageal spasm; HRM, high-resolution manometry; STP, spatiotemporal plot; LP, line plot by the AGA Institute /08/$34.00 doi: /j.cgh

2 526 GRÜBEL ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 5 of Melbourne Health. All participants gave informed consent. Students were given an incentive to perform, with the top 10 students awarded a book voucher. Manometric Data The manometric data to be used were retrieved from our database of clinical studies, with deidentified patient data. Swallows for both training and testing were selected to be normal or typical of each disorder and could be diagnosed accurately by either method according to conventional classification systems (ie, cases of focal dysmotility were excluded). 9 Our standard water-perfused manometric assembly contains 16 channels, with 8 channels at 1-cm intervals positioned across the LES and 8 channels at 3-cm intervals positioned in the esophageal body (Dentsleeve International, Mississauga, Canada). Data are collected at 25 Hz by using purpose-written software (Trace! G. Hebbard, written using LabVIEW; National Instruments Corp, Austin, TX), a locally constructed preamplifier system (Biomedical Engineering, The Royal Melbourne Hospital), 16-channel data acquisition card (National Instruments Corp), and a locally built perfusion pump. These highresolution data were displayed as a spatiotemporal color plot without contours, but with the color scale displayed on screen. LES data in the LP format were created by integrating data from 7 channels (over 6 cm) across the LES into a virtual sleeve by taking the highest pressure in each channel at any given time point. Data in the lowest recording channel were used as the gastric channel, and each channel in the body of the esophagus was displayed as a separate plot. The LP format therefore displayed 10 channels, 8 in the body of the esophagus, the LES virtual sleeve, and a gastric channel. Figure 1 shows a teaching screen of the electronic tutorial, with a normal swallow presented in both STP and LP. Training and Assessment Students registering an interest in the study were given a paper tutorial and an interactive compact disk covering esophageal physiology and a description of the 2 methods of data representation to study in their own time. The first part of the electronic teaching module contained identical information to the paper tutorial and discussed the anatomy and physiology of the esophagus, common esophageal motility disorders, and the diagnostic criteria for these. The second part of the electronic module consisted of an interactive tutorial, with data from the same swallow represented in both STP and LP formats, with the ability for students to move a cursor that controlled a display of the pressures at the relevant point in time/space (Figure 1). At the end of the electronic teaching module, students completed a short self-assessment section, with examples of 6 typical swallows (normal swallow, ineffective peristalsis, nutcracker esophagus, synchronous contraction, and achalasia) represented in STPs and LPs with the solution available to students while viewing the plot. This was in identical format to the evaluation (see below). The teaching module (paper and electronic) took approximately 90 minutes to complete. After completing training, students attended our department to sit an electronic assessment. The assessment consisted of data from 30 swallows (different swallows to those shown in the electronic tutorial) presented electronically in both LP and STP format (with a purpose-written program in LabVIEW; National Instruments Corp). The 60 (30 swallows in each representation) test screens were presented in fully randomized order (ie, the same swallow in its 2 representations could appear as screens 1 and 16 or 14 and 45 out of 60 screens). Pressures were measured by moving a cursor in an identical manner to that in the electronic tutorial. Participants were asked to evaluate each swallow for the adequacy of LES relaxation, presence and strength of peristalsis, and to make a diagnosis (Table 1). At all times they had access to a screen of reference values if required. After the 60 test screens, participants were asked to subjectively rate the ease of interpretation of data in each of the formats on a visual analogue scale and to give their preference for each of the methods of data presentation (Figure 2). Answers to the preference questions (Spatiotemporal Plot, Line Plot, or Equal) were presented in random order (newly randomized for each question for each subject). Figure 1. Teaching screen of electronic tutorial.

3 May 2008 VALUE OF SPATIOTEMPORAL REPRESENTATION OF MANOMETRIC DATA 527 Table 1. Classification Options LES relaxation Present Absent Esophageal peristalsis Present, normal amplitude Present, high amplitude Present, reduced amplitude Synchronous contraction Absent Diagnosis Normal swallow Nutcracker esophagus Ineffective peristalsis Diffuse esophageal spasm Achalasia Demographic Data Demographic data collected included sex, age, semester of medical course, previous education, prior exposure to manometry data and esophageal motor disorders, as well as each student s experience and comfort with data represented in graphical format and an estimate of the time that they spent with the printed and electronic tutorials. Times for the printed and electronic tutorials were added to give an overall study time. Data Analysis Manometry data. For each screen, the data recorded included the total time the screen was viewed by the subject and answers to each of the 3 multiple choice classification questions ([1] Peristalsis, [2] LES relaxation, and [3] Diagnosis). Scores for the number of correct answers were added for each subject, giving a score out of 180 overall or 90 for each data representation (30 screens for each representation and 3 classifications). Because swallows were paired, differences between STPs and LPs were examined by using Wilcoxon signed rank test for continuous data and McNemar test for dichotomous data. Linear regression was used to compare performance between the STPs and LPs. Subjective data. Visual analogue scores were converted to numeric data in the range of Differences in perceived difficulty of assessment between LP and STP were compared by paired t test, as data were normally distributed. Preferences (overall, difficulty, speed, accuracy, and suitability for doctors or patients) were analyzed by using loglinear modeling. All statistical analyses conducted were two-sided and were performed by using either StatView (SAS Institute, Inc, Cary, NC) or Stata v9 (Stata Corporation, College Station, TX) statistical software, with the significance level set as.05. Data are presented as mean standard error of the mean or median (25th 75th percentile). Results Sixty-nine tutorials were distributed; 60 students (27 women, 33 men; mean age, years; range, years) sat the electronic assessment. Fifty-nine were included in the analysis; 1 participant was excluded from analysis because this student achieved a score that was not distinguishable from random data entry. Overall study time (including the printed and electronic tutorial) was minutes. Before commencing the tutorial, the majority of the students were aware of (58%) but had not seen esophageal manometry (10%). The majority knew of the diagnosis of achalasia (80%), with smaller proportions recognizing diffuse esophageal spasm (DES) (42%), ineffective peristalsis (31%), or nutcracker esophagus (12%). Performance and Accuracy The mean number of total correct answers was 157, giving an overall accuracy of 87% 1.2%. There was a small but statistically significant difference in the total number of correct Figure 2. Questions of preference.

4 528 GRÜBEL ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 5 Table 3. Preferences Regarding the Plot Type STP LP No preference Overall preference 68% 12% 20% Faster to interpret 78% 15% 7% Easier to interpret 61% 19% 20% Better accuracy 58% 20% 22% Better suitability for doctors 58% 12% 30% Better suitability for patients 51% 41% 8% Figure 3. Number of total correct answers in both plots. answers in the STP representation compared with the LP representation (89% 1.2% vs 86% 1.3%, P.001) (Figure 3). A similar difference was found in the breakdown of the individual components of the score in terms of classification of peristalsis (STP vs LP, 84% 1.6% vs 81% 1.6%, respectively; P.005), LES relaxation (STP vs LP, 95% 1.0% vs 93% 1.3%, respectively; P.005), and diagnosis (STP vs LP, 86% 1.5% vs 80% 2.0%, respectively; P.007). The proportion of correct diagnoses by swallow type is detailed in Table 2. Differences between LP and STP scores were greatest in the group of students with the lowest third of LP scores, compared with those in the mid and upper tertiles (P.001). Time The mean time to analyze the data and complete the classifications for each screen was 28 seconds (range, seconds) overall, 26 seconds (range, seconds) for the STP, and 32 seconds (range, seconds) for the LP. The time spent viewing each screen was significantly less for the STP compared with the LP: mean time difference, 6 seconds; 95% confidence interval, 4 7; P.001. Correct answers were associated with shorter periods of time viewing the screen compared with incorrect answers (25 seconds [range, seconds] vs 41 seconds [range, seconds]; median difference, 11 seconds [range, 3 26 seconds]; P.001). The difference in time between STP and LP was significant both when the diagnosis was made correctly (mean time difference, 6 seconds; 95% confidence interval, 4 7; P.001), and when the diagnosis was incorrect (mean time difference, 5 seconds; 95% confidence interval, 4 7; P.001). The time each screen was viewed also related to the diagnosis. Students were quickest to correctly diagnose achalasia presented in the spatiotemporal format and were slowest to incorrectly diagnose a normal swallow presented in LP format. The time to make a correct diagnosis by swallow type is detailed in Table 2. Subjective Preference and Ratings The LP was rated more difficult to interpret than the STP (5.5 vs ; mean difference, 1.5; 95% confidence interval, ; P.001), and students subjectively preferred the spatiotemporal presentation (Table 3). With log-linear modeling, there was a strong preference for STP compared with LP (P values.001) for all comparisons (Table 3). Discussion We have shown that in manometry-naïve but medically oriented and intelligent students, after a short period of training, the spatiotemporal representation of manometry data was analyzed more quickly and more accurately than optimal LP representations of the same data. In addition, the students found the STPs subjectively easier to interpret and expressed a strong preference for this form of data presentation. They also considered that medical professionals and patients would find them easier to understand. Spatiotemporal presentation of data developed in parallel with HRM, 1 largely as a result of the complex data sets created by the large number of recording channels. When represented as conventional LPs, these data were cumbersome to display and analyze, yet they contained detailed information on the space-time structure of the esophageal pressure wave. The fact Table 2. Percent of Correct Answers and Time to Make a Correct Diagnosis % Correct answers Time to make a correct diagnosis (seconds) LP STP P value a LP STP P value a Achalasia 100 (73 100) 100 (73 100) (16 28) 16 (13 23).001 DES 73 (67 87) 73 (60 87) (24 50) 29 (22 47).24 Ineffective peristalsis 90 (83 97) 97 (87 100) (25 36) 24 (18 29).001 Normal 87 (80 100) 100 (87 100) (28 50) 26 (18 39).001 Nutcracker 100 (87 100) 100 ( ) (17 33) 17 (14 25).003 All types 90 (73 100) 97 (80 100) (21 40) 22 (16 32).001 NOTE. Data presented as median (25th 75th) percentile. a Wilcoxon signed rank test. With the Bonferroni correction, the adjusted significance level for multiple pairwise comparison was.01.

5 May 2008 VALUE OF SPATIOTEMPORAL REPRESENTATION OF MANOMETRIC DATA 529 that the data were digitized allowed novel representations to be used, and it was logical to represent the pressure as the dependent variable plotted against time and position along the manometric assembly, thus creating a map of pressure in space and time. This allows for an image-based evaluation (supplemented by simple measurement tools) that observers find easy to interpret because the human eye is highly attuned to image recognition. Clearly the potential for bias, either conscious or unconscious, is a significant issue in the interpretation of our findings. We attempted to minimize the effects of bias as much as possible in the design of the study. Training was conducted by commencing with general information on esophageal anatomy and physiology followed by an introduction to the LP and spatiotemporal methods of data representation and then examples and training on swallows in both formats. Students were told that the aim of the study was to compare different forms of data representation but were given no information as to why this was of interest. The same swallows were presented in the 2 formats in fully randomized order, and in addition to subjective data, objective measures of performance (accuracy and time spent analyzing and reporting data) were recorded. Data for the LPs were derived from the HRM data used for the STPs by creating a virtual sleeve to represent LES function as best practice LP data. As evidenced by the fact that the students performed well overall in both formats, they clearly acquired a good understanding of both techniques from the teaching module and were able to interact with the data to correctly classify swallows and achieve a diagnosis in nearly 90% of swallows with either type of data presentation. Students were approximately 20% faster when using the STPs than the LPs. Moreover, there was a marked preference for the spatiotemporal presentation of the data, indicating that they found this form of representation easier to understand. Students were faster in recognizing the diagnosis of achalasia and nutcracker esophagus than DES and ineffective peristalsis, indicating that students recognized the typical manometric findings of certain conditions more easily than others. These results are consistent with previous investigations of interobserver reproducibility both for LPs 7,10 and STPs, 7 in which diagnostic agreement was highest for normal peristalsis and achalasia and lowest for less well-defined ineffective or nonspecific motility disorders. In addition, students were faster in making a correct compared with an incorrect classification, suggesting that the majority used pattern recognition to reach diagnosis rather than a systematic approach to data analysis (that would be expected to improve with time spent in analysis). Students performed least well in recognizing the DES swallows and best in recognizing achalasia. For these swallows, there was no difference in performance between the LPs and STPs. The reasons for this are that all achalasic swallows were recognized correctly and, in the case of the DES swallows, might relate to the fact that the software did not contain a tool for measuring velocity. Interestingly, the group that benefited most in terms of accuracy from viewing the data in STP format were those who had the most difficulty classifying the LPs correctly, suggesting that the STP is a more intuitive and user-friendly method of data representation than a series of LPs. In this study, cases and swallows for both training and testing were selected to be normal or typical of each disorder that could be diagnosed accurately by either method according to conventional classification systems. 9 Previous comparisons between conventional manometric techniques with 4 6 LPs and HRM report diagnostic disagreement in 12% 20% of cases mainly as a result of increased yield of focal esophageal dysmotility and functional obstruction at the LES. 1,7,11 The case selection in this study increased the likelihood that the 2 techniques would be equivalent because the potential advantage of HRM is not fully tested here; rather, the primary aim was to compare interpretation of STPs and LPs. It is quite likely that if consecutive, unselected cases had been compared, the advantage of HRM would have been higher. Medical students were chosen as subjects who would be intelligent, physiologically and medically oriented, but with no previous experience or preconceptions regarding the representation of manometric data. We believe that this group can be reasonably considered to be representative of most gastroenterology trainees, the non-gastroenterologic medical population (and possibly general gastroenterologists who do not have a specific interest in esophageal motility) with whom we wish to communicate the significance of manometry findings. Our results indicate that the presentation of manometry data in spatiotemporal format is a more accessible form of data representation than the traditional series of LPs. Moreover, by improving understanding of difficult cases, especially in individuals who might have problems interpreting data presented in LP format, STP analysis should improve the interpretation of manometric data and understanding of esophageal physiology. References 1. Clouse RE, Prakash C. Topographic esophageal manometry: an emerging clinical and investigative approach. Dig Dis 2000;18: Nguyen NQ, Holloway RH. Recent developments in esophageal motor disorders. Curr Opin Gastroenterol 2005;21: Ghosh SK, Pandolfino JE, Zhang Q, et al. Quantifying esophageal peristalsis with high-resolution manometry: a study of 75 asymptomatic volunteers. Am J Physiol Gastrointest Liver Physiol 2006; 290:G988 G Dogan I, Mittal RK. Esophageal motor disorders: recent advances. Curr Opin Gastroenterol 2006;22: Clouse RE, Staiano A. Topography of normal and high-amplitude esophageal peristalsis. Am J Physiol 1993;265(Pt 1):G1098 G Clouse RE, Staiano A, Alrakawi A. Topographic analysis of esophageal double-peaked waves. Gastroenterology 2000;118: Fox M, Hebbard G, Janiak P, et al. High-resolution manometry predicts the success of oesophageal bolus transport and identifies clinically important abnormalities not detected by conventional manometry. Neurogastroenterol Motil 2004;16: Holloway RH. Topographical clinical esophageal manometry: a better mousetrap or manometric overkill? Am J Gastroenterol 2000;95: Spechler SJ, Castell DO. Classification of oesophageal motility abnormalities. Gut 2001;49: Nayar DS, Khandwala F, Achkar E, et al. Esophageal manometry: assessment of interpreter consistency. Clin Gastroenterol Hepatol 2005;3:

6 530 GRÜBEL ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No Fox MR, Bredenoord AJ. Oesophageal high-resolution manometry: moving from research into clinical practice. Gut 2008;57: Address requests for reprints to: Dr med Claudia Grübel, Kantonsspital St. Gallen, Department of Gastroenterology, Rorschacherstasse, 9007 St. Gallen, Switzerland. claudiagruebel@bluewin.ch; fax: Dr Grübel received a research grant from AstraZeneca Switzerland, Roche Pharma Switzerland, Altana Pharma Switzerland, and the Cancer Council of St. Gallen-Appenzell, Switzerland. Book vouchers for the highest scoring students were provided by AstraZeneca Australia. G.H. has been a member of advisory boards for and received travel support from AstraZeneca and Pfizer Australia. He also received research support from AstraZeneca. Ms Wendy Brown and Mr Paul Burton from Monash University helped with recruiting students. Agnes Dodds from the University of Melbourne assisted with student recruiting and completion of the questionnaires. Dr Mark Fox, University Hospital Zurich, assisted with critical reading of the manuscript.