Dr J. D. Gardner, Science for Organizations, Inc., 156 Terrace Drive, Chatham, NJ 07928, USA. E-mail: firstname.lastname@example.org
Background : In gastro-oesophageal reflux disease (GERD) subjects treated with a gastric anti-secretory agent, it is not known whether there is a relationship between heartburn severity and oesophageal acid exposure.
Methods : Oesophageal pH and heartburn severity were determined in 27 GERD subjects at baseline and on days 1, 2 and 8 of treatment with 20 mg omeprazole or 20 mg rabeprazole in a randomized, two-way crossover fashion.
Results : Receiver operating characteristic (ROC) analysis was used to determine values for heartburn severity that gave optimal cut-off points for distinguishing between normal and pathologic oesophageal reflux. Using these cut-off points, we found that the probability of no pathologic oesophageal reflux (Y) could be best fitted by an exponential equation: Y = a(e–bX) + c, where a, b and c are constants and X is the value of heartburn severity. There was close agreement between predicted and observed percentages of subjects with pathologic oesophageal reflux during different days of treatment.
Conclusions : In GERD subjects treated with a proton-pump inhibitor, the value of heartburn severity following a single standard meal can predict the likelihood of pathologic oesophageal reflux over the entire 24-h period.
In patients with gastro-oesopgheagal reflux disease (GERD), inhibition of gastric acid secretion with a proton-pump inhibitor or a histamine H2-receptor antagonist is the mainstay of treatment of oesophageal pathology and symptoms.1,2 On the other hand, in GERD subjects treated with a gastric anti-secretory agent, it is not known whether there is a relationship between heartburn severity and oesophageal acidity.
One reason for the lack of information regarding relationships between effects of anti-secretory treatment on gastric acid secretion or gastric acidity and effects on other features in GERD has been the lack of suitable analytical techniques. Previously3,4 we have described techniques using recordings of gastric and oesophageal pH that can quantify acidity over time. Two previous papers5,6 illustrated how these approaches can be used to demonstrate that meal-stimulated gastric acid secretion and post-prandial gastric acidity are significantly increased in GERD subjects, and the extent to which gastric acidity has to be decreased to prevent pathologic oesophageal reflux in GERD subjects who are being treated with a proton-pump inhibitor.
In the present study we analysed data from 24-h oesophageal pH recordings and measurements of heartburn severity following a standard meal in order to examine possible relationships between heartburn severity and oesophageal acidity. We found that there is a precise, quantitative relationship between heartburn severity following a single standard meal and the probability of pathologic oesophageal acid exposure during the entire 24-h period on the same day.
This study was approved by and conducted in compliance with good clinical practices as supervised by the Western Institutional Review Board, Olympia, WA. All subjects enrolled in this study gave written informed consent.
The analyses reported in this paper are based on the same data for oesophageal pH that were analysed in two preceding papers.5,6
GERD subjects were 59 adults with a history of GERD who experienced heartburn at least four times per week for at least 6 months (36 males, 23 females; ages 18–69 years). These subjects had no history of a serious medical condition, other clinically significant gastrointestinal illness (including peptic ulcer), difficulty in swallowing thought to be due to a pathologic process other than simple oesophageal reflux, history of a gastrointestinal haemorrhage, upper GI surgery, stricture or oesophageal dilation. Subjects had not been treated with an investigational drug within 30 days prior to study entry. Subjects had not used sucralfate, histamine H2-receptor antagonists, misoprostol, proton-pump inhibitors, pro-motility drugs or any other medication that alters acid secretion or gastrointestinal motility within 1 week of baseline pH recording. Subjects had not used systemic corticosteroids, anti-cholinergics, anti-neoplastic agents, metoclopramide, anticoagulants, tetracycline or cisapride within 1 month of baseline pH recording.
All subjects had a negative serology test for Helicobacter pylori.
The study was conducted at the Oklahoma Foundation for Digestive Research on the campus of the University of Oklahoma Health Sciences Center.
Baseline 24-h gastric pH and oesophageal pH were measured in 59 GERD subjects. Twenty-seven subjects with oesophageal pH ≤ 4 for at least 10% of the 24-h baseline recording period were randomized to receive eight consecutive daily doses of 20 mg omeprazole or 20 mg rabeprazole in a crossover fashion with a 14-day washout between treatment periods. Gastric pH and oesophageal pH were measured for 24 h on days 1, 2 and 8 with each treatment. The purpose of the present analyses is to assess effects of treatment with a proton-pump inhibitor, not to compare results with omeprazole to those with rabeprazole.
Subjects fasted from approximately 22:00 the evening before until the beginning of pH recording the following morning at 08:00. Standardized meals were provided at breakfast, lunch, dinner and bedtime. Smoking and ingestion of food other than the test meals were prohibited during the pH recording periods. The caloric content and amounts of fat, carbohydrate and protein in one portion of each meal were obtained from the providers of the meals. Subjects were allowed to ingest up to two portions of each meal, and the amounts of each meal ingested were recorded. Subjects were required to ingest the same amounts of food during all 24-h pH recording sessions. Meal 1 was ingested at 09:00 and consisted of one McDonald's sausage, egg, cheese, biscuit (bread roll) with 30 g onions, 8 oz chocolate milk and one Peppermint Patty. One portion contained 555 calories, 31 g fat, 43 g carbohydrate and 27 g protein. Meal 2 was ingested at 12:00 and consisted of one Wendy's 8 oz chili portion, one small frosty and 12 oz cola. One portion contained 670 calories, 15 g fat, 113 g carbohydrate and 23 g protein. Meal 3 was ingested at 18:00 and consisted of one Wendy's hamburger with cheese, one small portion of French fries and 12 oz cola. One portion contained 820 calories, 33 g fat, 108 g carbohydrate and 29 g protein. At bedtime (22:00) subjects ingested a snack consisting of two Hostess chocolate cupcakes and 8 oz whole milk.
During the evening meal (18:00) subjects rated heartburn severity on a 100-mm visual analogue scale (VAS scale) immediately prior to the meal and at 15-min intervals for 3 h beginning at the end of the meal. On the VAS scale, 0 indicated no heartburn and 100 indicated ‘heartburn as bad as it can get’.
Gastric pH and oesophageal pH values were collected using an ambulatory, dual-channel pH recording system (Medtronic Synectics) with antimony electrodes. One electrode was placed in the stomach 10 cm below and the second was placed in the oesophagus 5 cm above the manometrically defined upper border of the lower oesophageal sphincter. Electrodes were calibrated to pH 1 and 7, and connected to a portable data storage unit (Digitrapper, Medtronic Synectics). Recordings began at 08:00 and continued for 24 h. Data were transferred from the portable data-storage unit and processed using software designed for pH recordings (Polygram for Windows, Version 2.04, Medtronic Synectics). Only data from oesophageal pH measurements are analysed in the present report.
Polygram software takes the individual pH values that are recorded every 4 or 8 s and fills in the same value for the other seconds, resulting in one value for every second of the recording period for each electrode. These pH data were used for all calculations.
Integrated oesophageal acidity was calculated for each hour of the recording period as described previously.3,4,6
VAS scores were integrated by calculating the time-weighted average over the 3-h post-meal period. The maximum possible score is 300 points per hour and the minimum possible score is zero.
Statistical analyses were performed using GraphPad for InStat version 3.01 for Windows software. Curve-fitting was performed using GraphPad Prism version 3.00 for Windows.
In previous analyses of oesophageal pH recordings from 26 healthy subjects and the baseline pH recordings from the 59 GERD subjects in the present study, we found that a value of 8.1 mmol h/L for 24-h integrated oesophageal acidity gave maximal values for sensitivity and specificity in distinguishing GERD subjects from healthy subjects.6 We have defined pathologic oesophageal reflux as 24-h integrated oesophageal acidity greater than 8.1 mmol h/L. All subjects in the present study had pathological oesophageal reflux baseline in that 24-h integrated oesophageal acidity ranged from 10.7 to 120.4 mmol h/L.
For the present analyses, we first plotted values for heartburn severity versus the corresponding value for 24-h integrated oesophageal acidity. Figure 1 illustrates that, although there was a good deal of scatter in the data, there was a significant correlation (P < 0.0001) between values for heartburn severity and corresponding values for integrated oesophageal acidity. At least some of the variation illustrated in Figure 1 may reflect within-subject variation, which we did not assess in the present study. The Spearman correlation coefficient and 95% confidence interval were 0.375 (0.238, 0.497). This correlation coefficient of 0.375 indicates that even though there was statistically significant relationship between heartburn severity and the corresponding value for 24-h integrated oesophageal acidity, the overall correlation was poor.
The data in Figure 1 could be fitted by a straight line with the equation
where Y is heartburn severity and X is oesophageal acidity. Solving this equation for integrated oesophageal acidity of 8.1 mmol h/L gives a cut-off point of heartburn severity of 56.7. In Figure 1, 83 values for heartburn severity were higher than 56.7. Of these 40 (48%) had corresponding values of integrated oesophageal acidity that were higher than 8.1 mmol h/L, indicating pathologic oesophageal reflux. Ninety-nine values for heartburn severity were less than 56.7 and of these, 26 (26%) had corresponding values for integrated oesophageal acidity that were higher than 8.1 mmol h/L. These results indicate that using heartburn severity of 56.7 as a cut-off point has a sensitivity of 48% and a specificity of 74% to distinguish between subjects with and without pathologic oesophageal reflux. The wide scatter in the data in Figure 1 and the low correlation coefficient caused us to consider another approach.
We next examined the ability of heartburn severity to distinguish between subjects with and without pathologic oesophageal reflux by performing a receiver operation characteristic (ROC) analysis.7–10Figure 2-left shows the percentage of subjects with pathologic oesophageal or normal oesophageal acid who had values for heartburn severity above the value indicated on the x-axis of the graph. Values from subjects with pathologic oesophageal acid were significantly higher than values from subjects with normal oesophageal acid (P < 0.0001 by the Wilcoxon signed-rank test). Figure 2-right is a ROC7 curve of the data in Figure 2-left, and examines the extent to which values for heartburn severity differ between subjects with and without pathologic oesophageal acid. This curve plots the sensitivity (i.e. true positives or percentage of subjects with pathologic oesophageal acid) versus 1-specificity (i.e. false positives or percentage of subjects with normal oesophageal acid) for each value of heartburn severity. In Figure 2-right, the more the points are displaced from the diagonal identity line, the more heartburn severity distinguishes between subjects with and without pathologic oesophageal acid. The area under the ROC curve, which is conceptually equivalent to the Wilcoxon statistic,8–10 was 0.675, and represents the probability that a subject selected randomly from the group with pathologic oesophageal acid will have a higher value for heartburn severity than a subject selected randomly from the group with normal oesophageal acid.7–10 In Figure 2 the optimal cut-off point for heartburn severity that distinguished between subjects with and without pathologic oesophageal acid was 93.8 points per h. This value had a sensitivity of 57% and a specificity of 91%.
Sensitivity and specificity assess how well a particular value of heartburn severity distinguishes between subjects with and without pathologic oesophageal reflux. That is, given one group of subjects with pathologic oesophageal reflux and another group without pathologic oesophageal reflux, sensitivity and specificity assess how well heartburn severity can distinguish between these two groups. In other words, sensitivity and specificity are results from analyses that go from the groups to an individual value.8,11 What sensitivity and specificity do not address, however, is the ability of a value for heartburn severity in an individual subject to predict the likelihood of pathologic oesophageal acid. The positive predictive value addresses the question ‘If a subject has a value above the cut-off point for heartburn severity, what is the probability that the subject has pathologic oesophageal reflux?’ and the negative predictive value addresses the question ‘If a subject has a value below the cut-off point, what is the probability that the subject does not have pathologic oesophageal reflux?’ In other words, the positive and negative predictive values are results from analyses that go from an individual value to the group. An important feature of the predictive values is that they depend on the prevalence of pathologic oesophageal reflux in the sample, and the lower its prevalence, the is lower the ratio of true positives to false positives. (A particularly clear discussion is given in ref. 11.)
We next analysed the data used for the analyses illustrated in Figure 2 to determine the predictive value, i.e. to calculate the probability that a subject with a given value for heartburn severity has no pathologic oesophageal reflux. We calculated the probability of no pathologic reflux as the percentage of records with integrated oesophageal acidity below 8.1 mmol h/L at each value of heartburn severity. In Figure 3 the probability of no pathologic reflux is 100% when heartburn severity is near zero, and with increasing values of heartburn severity, the probability decreases progressively to 63%. As mentioned above, values for the probability of no pathologic reflux depend on the percentage of records without pathologic oesophageal reflux. In the case of no pathologic reflux, defined as integrated oesophageal acidity below 8.1 mmol h/L, 63% of the records used for the present analyses had no pathologic reflux; therefore, the probability of no pathologic reflux can never be less than 63%.
The results in Figure 3 were best fitted by an equation with a single exponential plus a plateau, i.e.
where Y is the probability of no pathologic reflux, k is a constant with a value of 8.922 × 10−2, and X is the value for integrated heartburn severity. This equation fitted the data in Figure 3 with R2 of 0.892, indicating that virtually all of the variation in the probability of no pathologic reflux could be accounted for by values for heartburn severity.
Figure 4 compares the calculated probability of no pathologic oesophageal reflux using the equation above and the observed percentage of subjects with no pathologic oesophageal reflux days 1, 2 and 8 of treatment with a proton-pump inhibitor. There was close agreement between calculated and observed values on days 2 and 8; however, on day 1, the predicted probability was higher than the observed value. These differences on day 1 occurred because the predicted value for no pathologic oesophageal reflux for a given subset of the population can never be less than the prevalence of no pathologic oesophageal reflux for the entire population, which in the present study was 63%. In addition, some subjects may have had oesophageal reflux before measurement of heartburn severity because of the delayed onset of action of the first dose of the proton-pump inhibitor. That is, even though on day 1, subjects have decreased heartburn severity with the evening meal, reflecting the action of the proton-pump inhibitor on oesophageal acidity, some subjects probably experienced oesophageal reflux earlier in the day before the onset of action of the drug.
The main purpose of the present report was to examine the extent to which heartburn severity after a single standard meal might be related to oesophageal acidity over the same 24-h period in GERD subjects being treated with a proton-pump inhibitor. It is well established that in GERD subjects, proton-pump inhibitors reduce the frequency and severity of heartburn and decrease oesophageal as well as gastric acidity.1,2,12 We are unaware, however, of studies that have examined possible relationships between heartburn severity or frequency and oesophageal acidity in individual GERD subjects. In the present study, we found that in GERD subjects treated with a proton-pump inhibitor, (1) heartburn severity was significantly higher in subjects with pathologic oesophageal reflux than in subjects without pathologic reflux, (2) heartburn severity predicted the likelihood of pathologic oesophageal reflux with high precision, and (3) after the first day of treatment, there was good agreement between the observed percentage of subjects with pathologic oesophageal reflux and the percentage predicted by the subjects' heartburn severity.
Our primary focus in the present study was not on the value of oesophageal acidity per se, but on whether or not the value was pathologic using a cut-off point based on values for 24-h integrated oesophageal acidity in normal and GERD subjects.6 In this previous study, we defined pathologic oesophageal reflux in the same way as others13,14 using the value for oesophageal acid exposure calculated from 24-h oesophageal pH recordings that best distinguished between GERD subjects and normal subjects in terms of sensitivity and specificity.6 Under other circumstances, other criteria for establishing cut-off points might be used. For example, a different value might be used to distinguish between GERD subjects with and without erosive oesophagitis, while still another value might be used to distinguish between GERD subjects with and without resolved heartburn.
We found that there was a significant, direct correlation between heartburn severity after a single meal and integrated oesophageal acidity over the same 24-h period. It was possible using a least-squares regression analysis to calculate the value of heartburn severity that corresponded to the cut-off point for oesophageal acidity that distinguished normal from pathologic oesophageal reflux. It was then possible to use this cut-off point to determine the sensitivity and specificity with which this value for heartburn severity distinguished between normal and pathologic oesophageal reflux. The problem, however, was that there was no way to know that this value for heartburn severity provided optimal discrimination between normal and pathologic oesophageal reflux.
To determine the value of heartburn severity that had the highest sensitivity and specificity for distinguishing between normal and pathologic oesophageal reflux, we employed a ROC analysis. This analysis indicated that the value of heartburn severity which provided optimal sensitivity and specificity for pathologic oesophageal reflux was approximately two times higher than the value determined from least-squares regression analysis. The ROC analysis also indicated that values for heartburn severity from subjects with pathologic oesophageal reflux were significantly higher than values from subjects without pathologic reflux. The area under the ROC curve indicated that the probability that a subject chosen randomly from the group with pathologic oesophageal reflux would have a higher value for heartburn severity than a subject chosen randomly from the group without pathologic reflux was 0.675.
We found that the probability of pathologic oesophageal reflux in subjects treated with a proton-pump inhibitor could be predicted with high precision using an equation with a single exponential plus a plateau that reflected the percentage of the total records with no pathologic oesophageal reflux. There is no mechanistic or pathophysiologic significance to this equation of which we are aware. It is simply the equation that best fits the data. We also found that there was close agreement between the observed and predicted percentages of subjects with no pathologic oesophageal reflux on days 2 and 8 of treatment. As mentioned above, the difference between observed and predicted percentages on day 1 occurred because the probability of no pathologic oesophageal reflux on day 1 can never be less that the overall prevalence of no pathologic oesophageal reflux.
There are several caveats related to the present analyses. First, the results are probably only applicable to GERD subjects treated with a proton-pump inhibitor. Similar subjects treated with a histamine H2-receptor antagonist would likely give different results because these agents have a different duration of action on gastric acid secretion, have different effects on gastric pH during fasting, and inhibit fasting gastric acid secretion to a greater extent than meal-stimulated secretion.15–17 Second, in the present study, the GERD subjects treated with a proton-pump inhibitor were those with oesophageal pH ≤ 4 for at least 10% of a 24-h baseline recording period. The parameter values for the exponential equation that gives the probability of no pathologic reflux might differ in important ways with GERD subjects who have lower values of oesophageal acidity. Third, we3 and others18 have reported that GERD subjects with erosive oesophagitis have higher values for oesophageal acid exposure than GERD subjects without erosive oesophagitis. It may be that the relationship between heartburn severity and oesophageal acidity during treatment with a proton-pump inhibitor differs between these two groups. Fourth, we have found that GERD subjects can be classified in terms of whether most of their oesophageal acid exposure occurs before or after 23:00.3,4 The relationship between heartburn severity and oesophageal acidity during treatment with a proton-pump inhibitor might differ between these two groups.
The present study demonstrates that there is a precise quantitative relationship between heartburn severity after a single standard meal and the likelihood of pathologic oesophageal reflux during the entire 24-h period. There have been no previous demonstrations of a relationship between heartburn severity or frequency and oesophageal acid exposure, probably because of the lack of suitable techniques for quantifying oesophageal acidity. Traditionally, time with pH ≤ 4 has been used to assess oesophageal as well as gastric acidity.1,13,14 Previously, however, we have documented some of the shortcomings of time with pH ≤ 4 compared to integrated acidity in evaluating gastric and oesophageal pH recordings.3,4,6 In the present study, we did not use time with oesophageal pH ≤ 4 to assess pathologic oesophageal reflux, because we found previously that time with oesophageal pH ≤ 4 is less sensitive than integrated oesophageal acidity in assessing the relationship between integrated gastric acidity and the probability of no pathologic oesophageal reflux.
In GERD subjects with pathologic oesophageal reflux, one can use the parameter values for the exponential equation to predict the probability of normalising oesophageal acid exposure during treatment with a proton-pump inhibitor from the corresponding value for heartburn severity after a standard meal. We realise that it might not be practical for clinicians to assess heartburn severity every 15 min after a standard meal; however, future studies might demonstrate that categorical assessments (such as none, mild, moderate and severe at hourly intervals) can provide satisfactory prediction of pathologic oesophageal reflux.
This work was supported by grants from Janssen Pharmaceutica Research Foundation to the Oklahoma Foundation for Digestive Research and to Science for Organizations, Inc.