Can esophageal pH monitoring predict delayed gastric emptying?

Can Esophageal pH Monitoring Predict Delayed Gastric Emptying? By Jose´ Esteva˜o-Costa, Jorge Amil Dias, Miguel Campos, Eunice Trindade, Armando Teixeira-Pinto, and Jose´ Luis Carvalho Porto, Portugal

Background/Purpose: Delayed gastric emptying (DGE) is frequent in patients with gastroesophageal reflux disease (GERD) and may require additional investigation. The current study assesses whether relative esophageal exposure, postprandial (PP) versus fasting, diagnosed by pH monitoring could predict DGE. Methods: Thirty patients with GERD underwent extended esophageal pH monitoring and were assigned as DGE or non-DGE according to scintigraphy. The PP to fasting ratio for reflux index, relative frequency of long episodes in PP, and distribution of the longest episode were used to assess the relative esophageal exposure. The effectiveness of these parameters to predict DGE was estimated; the cutoffs for continuous variables were chosen with receiver operating characteristics (ROC) curves and the probabilities were calculated using a logistic regression model. Results: The area under the ROC curve of PP to fasting ratio for reflux index was greater than that of relative frequency of long episodes in PP. There was a good equilibrium between

D

ELAYED GASTRIC EMPTYING (DGE) has been documented in a high proportion of patients with gastroesophageal reflux disease (GERD).1-5 Poor gastric emptying is probably partially responsible for postoperative gas-bloat syndrome and for failure of antireflux procedures owing to breakdown or slipping of the fundic wrap.6 Therapeutic enhancement of gastric emptying, however, has been used with apparent success in GERD with DGE7,8; actually adding a pyloroplasty to fundoplication seems to reduce the postoperative recurrence of GERD.8 These features dictate the necessity of additional investigation to assess gastric emptying, especially in patients who are candidates for surgical treatment.9 Measurement of gastric emptying is therefore important to not only the physiologist investigating gastric motility but also the clinician evaluating gastroduodenal dysfunction. Gastric scintigraphy currently is the gold standard to assess gastric emptying but is a time-consuming study not often readily available and involves radiation hazards for patients and personnel.10 These make it a demanding not ubiquitously available investigation not recommended for repeated examinations. We previously have shown that children with DGE present a relative esophageal exposure (postprandial-PP v fasting) significantly different from that found in those with non-DGE.11 Therefore, we hypothesized that pH Journal of Pediatric Surgery, Vol 39, No 10 (October), 2004: pp 1537-1540

sensitivity and specificity at a PP to fasting ratio of 1. A PP to fasting ratio greater than 1, ie, a reflux index in PP greater than in fasting, presented a sensitivity of 93% and a negative predictive value of 91%. The occurrence of the longest episode in PP had a specificity of 94% and a positive predictive value of 89%. A reflux index greater in PP plus a longest episode in PP presented a 94% probability of DGE; a reflux index greater in fasting plus a longest episode in fasting had a 95% probability of non-DGE. These combinations represented 60% of the series. Conclusions: PP to fasting ratio for reflux index and distribution of the longest episode seem accurate to identify DGE; thus, additional investigation to assess gastric emptying may be avoided in the majority of patients. J Pediatr Surg 39:1537-1540. © 2004 Elsevier Inc. All rights reserved. INDEX WORDS: Gastroesophageal reflux, delayed gastric emptying, scintigraphy, esophageal pH monitoring.

monitoring could predict DGE, avoiding additional investigation. The current study aims to evaluate the effectiveness of derived esophageal pH monitoring parameters to diagnose the type of gastric emptying. MATERIALS AND METHODS Thirty children with GERD were included in a prospective protocol. It encompassed candidates for surgical treatment consecutively evaluated between June 1995 and May 1999 in whom medical treatment has failed. Extended esophageal pH monitoring and gastric emptying scintigraphy were performed according to methodology previously published.11 Briefly, on pH monitoring a drop in pH below 4 lasting at least 20 seconds was considered a reflux episode; the PP period was defined as 120 minutes after a meal and the fasting period as all other times of the day; the following data were analyzed: reflux index (fraction of time when pH was less than 4), number of reflux episodes per hour, number of long (more than 5 minutes) episodes, and duration of the longest episode. Scintigraphic studies were performed after ingestion of a standardized semisolid meal (300 mL plus 500 ␮Ci99mTc/m2 body

From the Division of Pediatric Surgery, Unit of Pediatric Gastroenterology, and Department of Biostatistics and Medical Informatics, Faculty of Medicine, Hospital Sa˜o Joa˜o, Porto, Portugal. Address reprint requests to Professor J. Esteva˜o-Costa, Servic¸o Ciru´rgica Pediatria, Faculdade Medicina Porto, Hospital Sa˜o Joa˜o, 4200-319 Porto, Portugal. © 2004 Elsevier Inc. All rights reserved. 0022-3468/04/3910-0017$30.00/0 doi:10.1016/j.jpedsurg.2004.06.023 1537

ESTEVA˜O-COSTA ET AL

1538

Table 1. Clinical Data, Investigational Data, and Relative Esophageal Exposure Diagnosed by pH Monitoring (PP v Fasting)

Gastric Emptying (t ⁄ ) Age (yr) Sex (male/female) Esophagitis Global pH monitoring Total duration min PP/fasting Reflux index PP fasting Reflux episodes per hour PP fasting PP/fasting ratios Reflux index Reflux episodes per hour Relative frequency of long episodes in PP Longest episode in PP (%) 12

DGE (n ⫽ 14)

Non-DGE (n ⫽ 16)

102 [89-125.5] 3.5 [2-4.7] 10/4 29%

58.5 [45.5-75.5] 3 [1.7-11] 11/5 38%

.45* .59† .44†

1235 [1195-1284] 0.64 [0.49-0.69]

1241 [1174-1267] 0.69 [0.44-0.87]

.91* .49*

18 [9.9-32.2] 8.5 [4.5-16.5]

27.6 [10.7-60] 23.9 [15.6-36.8]

.49* .01*

13.5 [4.8-28.3] 8.5 [3.9-15]

11.6 [7-17.3] 7.1 [5.9-10.2]

.57* .95*

2.1 [1.4-2.6] 1.8 [1.5-2.1] 63 [50-80] 57

0.9 [0.8-1.5] 1.8 [0.8-2.2] 38 [32-60] 6

.002* .62* .04* .003†

P Values

NOTE. Numbers in square brackets are the first and third quartiles of the preceding variable (median). *Two-tail U Mann-Whitney test. †Fisher’s Exact test.

surface area); the radionuclide counts were obtained every 2 minutes for a recording period of 120 minutes over the stomach; DGE was defined as a t ⁄ greater than 85 minutes. Sixteen patients presented non-DGE, and 14 had DGE. Relevant clinical and investigational data are summarized in Table 1. The relative esophageal exposure (PP v fasting) was evaluated through PP to fasting ratios for reflux index and for number of reflux episodes per hour, relative frequency of long episodes in PP (PP ⫻ 100/PP ⫹ fasting) and distribution of occurrences for the longest episode. The latter parameter refers to the proportion of patients in each group (DGE v non-DGE) that presented the longest episode in PP versus in fasting. Children with DGE presented values of PP to fasting ratio for reflux index, relative frequency of long episodes in PP, and occurrence of the longest episode in PP significantly higher than patients with non-DGE (Table 1). The effectiveness (sensitivity, specificity, positive and negative predictive values) of these derived pH monitoring parameters was calculated for DGE. The cutoff values for continuous variables that were used to create a rule of classification for DGE or non-DGE were chosen based on the receiver operating characteristic (ROC) curves. The ROC curve is a plot of the sensitivity and specificity of each classification rule based on all possible cutoff values of the predictive variable.12 A multivariate logistic regression was used to compute the probability of DGE combining the longest episode in PP versus in fasting and PP to fasting ratio13,14; additionally, the same probabilities were computed by stratification, ie, directly from the data. The Hosmer and Lemeshow test was used to assess the fit of the logistic regression model.14 Statistical analysis was performed using SPSS v.9. The protocol for the study was approved by the ethics committee of our hospital, and informed consent was obtained. 12

val: 0.66 to 0.98) was greater than that of the latter (area ⫽ 0.71, 95% confidence interval: 0.52 to 0.89). Thus, only the PP to fasting ratio for reflux index was used; there was a good equilibrium between sensitivity (86% to 93%) and specificity (63%) for the cutoff of 1 (Fig 1). The sensitivity and the negative predictive value of PP to fasting ratio greater than 1 (ie, a reflux index in PP greater than in fasting) were high; the specificity and the

RESULTS

The ROC curve of PP to fasting ratio for reflux index was almost always greater than the one of relative frequency of long episodes in PP; the area under the ROC curve of the former (area ⫽ 0.82, 95% confidence inter-

Fig 1. ROC curves of PP to fasting ratio for reflux index and of relative frequency of long episodes in PP. The PP to fasting ratio 1 is situated between A (0.975) and B (1.155).

PREDICTING GASTRIC EMPTYING BY pH MONITORING

1539

Table 2. Effectiveness of Derived pH Monitoring Parameters to Diagnose DGE

PP/fasting ratio ⬎1* Longest episode in PP

Sensitivity

Specificity

Positive Predictive Value

Negative Predictive Value

93% (66%-100%) 57% (29%-82%)

63% (35%-85%) 94% (66%-100%)

68% (43%-87%) 89% (52%-100%)

91% (59%-100%) 71% (48%-89%)

NOTE. 95% confidence intervals in parentheses. *Reflux index greater in PP.

positive predictive value of occurrence of the longest episode in PP were also high (Table 2). The probability of DGE (Table 3) in the presence of both parameters (PP to fasting ratio ⬎1, ie, reflux index greater in PP, plus longest episode in PP) was very high; in the absence of both parameters (PP to fasting ratio ⬍1, ie, reflux index greater in fasting, plus longest episode in fasting) the probability of DGE was very low, that is, the probability of Non-DGE was very high. These 2 combinations represented 60% of the series (Table 3). The logistic regression model used to compute these probabilities did not show evidence of lack of fitness (P ⫽ .382). DISCUSSION

We have shown previously that when comparing the individual variation between PP and fasting exposures on pH monitoring, the patients with DGE present a relative esophageal exposure significantly different from that found in patients with non-DGE.11 Briefly, the patients with DGE compared with those with non-DGE presented a reflux index significantly lower in fasting but similar in PP, which resulted in a significantly higher PP to fasting ratio; moreover, the former had a relative frequency of long episodes in PP and an occurrence of the longest episode in PP significantly higher, ie, DGE accentuates the PP esophageal exposure probably by increasing the volume of refluxate per episode of reflux just as the PP to fasting ratio for reflux episodes per hour was similar to the one in the Non-DGE group.11 Thus, we hypothesized that those aforementioned derived parameters could be Table 3. Prevalence of Possible Combinations in Current Series and Respective Probabilities of DGE Computed by the Multivariate Logistic Regression Model

PP/fasting ratio ⬎1*

PP/fasting ratio ⬍1†

Longest Episode in PP

Longest Episode in Fasting

Prevalence: 26.7% Probability: 94% (8/8) Prevalence: 3.3%

Prevalence: 36.7% Probability: 50%

Probability: 47%

Prevalence: 33.3% Probability: 5% (1/10)

NOTE. Probabilities computed directly from data in parentheses (n/N). *Reflux index greater in PP. †Reflux index greater in fasting.

useful in assessing the type of gastric emptying if its effectiveness to diagnose DGE was satisfactory. The compromise between sensitivity and specificity of PP to fasting ratio for reflux index was more adequate than that of the relative frequency of long episodes in PP. At a PP to fasting ratio for reflux index of 1, a cutoff easy to apply in clinical practice, there was a good equilibrium between sensitivity and specificity. The analysis of effectiveness for the diagnosis of DGE found a high sensitivity and a high negative predictive value, ie, a positive result (PP to fasting ratio ⬎1, reflux index greater in PP) identifies the patients with DGE and a negative result (PP to fasting ratio ⬍1, reflux index greater in fasting) excludes DGE. The analysis of effectiveness of the distribution of occurrence of the longest episode showed a high specificity and a high positive predictive value, ie, a negative result (longest episode in fasting) identifies patients without DGE and a positive result (longest episode in PP) confirms DGE. Consequently, it seemed logical to associate the complementary effectiveness of both tests. The calculated probabilities of DGE in the presence (positive results) and absence (negative results) of both parameters, combinations that were found in the majority of patients, were satisfactorily high and low, respectively, to propose the methodology shown in Table 4. The current study has some theoretical limitations. In fact, the reproducibility was not ensured for all subsets of GER patients; however, surgical candidates are the patients that, in practice, may obtain the most benefit from evaluation of gastric emptying before treatment.8 One must realize, however, that the cutoffs were not tested on a different group of patients; despite the good concordance between the gold standard and the new method, it is important to verify the decision rule on a new cohort of patients. The derived esophageal pH monitoring parameters (PP to fasting ratio for reflux index and distribution of Table 4. Proposed Methodology to Predict Gastric Emptying Through Esophageal pH Monitoring

PP/fasting ratio ⬎1* PP/fasting ratio ⬍1†

Longest episode in PP

Longest episode in fasting

DGE Inconclusive

Inconclusive Non-DGE

*Reflux index greater in PP. †Reflux index greater in fasting.

ESTEVA˜O-COSTA ET AL

1540

the longest episode) seem to be a simple and accurate method to identify DGE; thus, additional studies to

assess gastric emptying may be avoided in a significant proportion of the patients.

REFERENCES 1. Cunningham KM, Horowitz M, Riddell PS, et al: Relations among autonomic nerve dysfunction, oesophageal motility, and gastric emptying in gastro-oesophageal reflux disease. Gut 32:1436-1440, 1991 2. Soykan I, Lin Z, Jones S, et al: Gastric myoelectrical activity, gastric emptying and correlations with dyspepsia symptoms in patients with gastroesophageal reflux. J Investig Med 45:483-487, 1997 3. Velasco N, Hill LD, Gannan RM, et al: Gastric emptying and gastroesophageal reflux: Effects of surgery and correlation with esophageal motor function. Am J Surg 144:58-61, 1982 4. McCallum RW, Berkowitz DB, Lerner E: Gastric emptying in patients with gastroesophageal reflux. Gastroenterology 80:285-291, 1981 5. Benini L, Sembenini C, Castellani G, et al: Gastric emptying and dyspeptic symptoms in patients with gastroesophageal reflux. Am J Gastroenterol 91:1351-1354, 1996 6. Borgstein ES, Heij HA, Beugelaar JD, et al: Risks and benefits of antireflux operations in neurologically impaired children. Eur J Pediatr 53:248-251, 1994 7. Carroccio A, Iacono G, LiVoti GL, et al: Gastric emptying in

infants with gastroesophageal reflux: Ultrasound evaluation before and after cisapride administration. Scand J Gastroenterology 27:799-804, 1992 8. Bustorff-Silva J, Fonkalsrud EW, Perez CA, et al: Gastric emptying procedures decrease the risk of postoperative recurrent reflux in children with delayed gastric emptying. J Pediatr Surg 34:79-83, 1999 9. Richter JE: Delayed gastric emptying in reflux patients: To be or not to be? Am J Gastroenterol 92:1077-1078, 1997 10. Elashoff JD, Reedy TJ, Meyer JH: Analysis of gastric emptying data. Gastroenterology 83:1306-1312, 1982 11. Esteva˜o-Costa J, Campos M, Dias JA, et al: Delayed gastric emptying and gastroesophageal reflux: A pathophysiologic relationship. J Pediatr Gastroenterol Nutr 32:471-474, 2001 12. Altman DG, Bland JM: Diagnostic tests 3: Receiver operating characteristics plots. BMJ 309:188, 1994 13. Fletcher RM, Fletcher SW, Wagner EH: Clinical Epidemiology: The Essentials. (ed 3). Baltimore, MD, William & Wilkins, 1996 14. Hosmer D, Lemeshow S: Applied Logistic Regression. (ed 1). Massachusetts, John Wiley & Sons, 1989