Recurrent abdominal pain (RAP) is the most common chronic pain entity in children and adolescents, affecting an estimated 10% to 15% of school-age children at any time (1). Apley and Naish provided a description of RAP in the late 1950s that, in retrospect, probably included children with both functional and organic causes for pain (2). Because RAP was a description of a heterogeneous group of disorders with varying degrees of symptom frequency, intensity and associated disability, the label of RAP did not provide a guide for the efficient evaluation or treatment of these children. Over time, RAP became a diagnosis applied to children with all kinds of nonorganic abdominal pain. Many tests were required to rule out organic causes for pain before the diagnosis could be applied. At the end of a negative medical evaluation, many clinicians and parents considered RAP to be a diagnosis implying the physician's belief that the complaint was imaginary or psychological in origin.
In recent years, there has been an effort to redefine RAP in ways that can better guide diagnostic evaluation and treatment while minimizing the use of invasive testing. Continuing a process begun in 1990, the Rome II pediatric working team in 1999 established discrete diagnostic criteria for several functional gastrointestinal disorders (FGIDs), including conditions with chronic non-organic abdominal pain (3). The pain-related diagnoses include irritable bowel syndrome (IBS), functional dyspepsia (FD), functional abdominal pain syndrome (FAPS) and abdominal migraine. Rome II diagnostic categories are based on specific clusters of gastrointestinal symptoms in the absence of easily identified organic disease likely to account for symptoms. Because Rome II diagnoses are symptom-based, they are diagnoses of inclusion rather than exclusion. Walker et al. found that 73% of patients referred for evaluation of RAP fulfilled Rome II criteria for an FGID by parental reports (4). Thus, Rome II criteria appear to capture the majority of children with RAP.
The utility of the Rome II criteria in clinical practice (as a guide for treatment) or in research (as a meaningful classification of participants) has not been determined. This study examined the frequency with which a Rome II diagnosis is made in children/adolescents referred for evaluation of abdominal pain from the perspective of the child, the parent and the diagnosing physician. We examined the rate of parent, child and physician agreement in assigning a specific Rome II diagnosis. We also analyzed the discrepancies between parent, child and physician evaluations to determine whether modifications to the Rome II criteria are needed.
PATIENTS AND METHODS
We enrolled 154 children/adolescents and their primary caregivers referred to a pediatric gastroenterology clinic for evaluation of chronic or recurrent abdominal pain. Families were approached for enrollment if the child was between 8 and 18 years of age and had abdominal pain of ≥12 weeks duration. Patients with known organic disease, previous abdominal surgery or significant chronic illness were excluded. In children who had endoscopy after their first visit, organic diseases excluding participation were Crohn's disease (1), giardiasis (1), eosinophilic esophagitis (2) and eosinophilic gastroenteritis diagnosed histologically (>40 eosinophils/high power field; 2). These diagnoses were excluded because the inflammation involved is generally accepted as causative of symptoms and, thus, would exclude them under Rome II criteria. Patients with visually normal endoscopies (including those whose biopsies showed mild-to-moderate histologic inflammation) were considered to have functional gastrointestinal disorders and were not excluded. The clinical significance of mild-to-moderate inflammation remains unknown. We felt that mild or moderate histologic inflammation did not specifically exclude these patients from a Rome II diagnosis.
Approximately 75% of the families approached agreed to participate. Participants were predominantly female (64%) and white (85%). Mean age of participants was 11.8 ± 2.5 years (range, 8 to 18 years).
The Questionnaire on Pediatric Gastrointestinal Symptoms (QPGS) assesses symptoms associated with FGIDs and provides specific scoring instructions based on the Rome II criteria to obtain an FGID diagnosis (5). There are separate questionnaires for children and parents. Recent research efforts have used the QPGS to classify children by FGID subtype (4).
During the first visit, the child and the primary caregiver independently completed the QPGS. A board-certified pediatric gastroenterologist then conducted a standardized complete history and physical examination. On the basis of the history/physical and blinded as to the QPGS results, the physician decided whether the child's symptoms fulfilled criteria for an FGID based on Rome II criteria. When diagnoses were discrepant, a single clinician (CF) evaluated each child and parent QPGS and the physician diagnostic summary to determine the reasons for discrepant diagnoses.
All procedures involved in this study were approved by the Institutional Review Boards of the participating hospitals.
Statistical comparisons were made using Pearson χ2 tests with significance accepted at P < 0.05. Fisher exact tests were used as an alternative when at least one cell size was less than 5. Kappa and weighted kappa were used to examine the agreement rate for diagnoses of dyspepsia versus IBS for all dyads (e.g., parent-child, child-physician).
Frequency of Functional Gastrointestinal Disorders
Of 154 children referred for RAP, 6 (4%) were excluded when an organic diagnosis was made. The frequencies of the Rome II diagnoses made by parent report, child report and physician evaluation for the remaining 148 children are shown in Figure 1. A large proportion of these children met criteria for an FGID based on the Rome II criteria by child report (84%), parent report (89%) and physician clinical impression (100%). Functional dyspepsia (FD) was the most common diagnosis, comprising 35% by child report, 47% by parent report and 57% by physician evaluation. For children diagnosed with FD, pain was the predominant symptom in 77% by child report and in 74% by parent report. IBS was diagnosed in 30% by child report, 20% by parent report and 12% by physician evaluation. Symptom-based criteria for both FD and IBS were met by 10% of children by their own report, 10% by parent report and 30% by physician evaluation. Functional abdominal pain syndrome and abdominal migraine were diagnosed in less than 10% of patients by any source.
Reasons for “No Diagnosis”
Eleven percent of children did not meet FGID criteria by parent report and 16% did not meet FGID criteria by child report. Reasons for not meeting criteria are shown in Table 1. Duration of symptoms was more likely to be a factor for “no diagnosis” for child report (58%) than for parent report (22%) (P < 0.05, Fisher exact test). There were no other significant differences between parents and children in reasons for “no diagnosis.”
Parent report led to a diagnosis of dyspepsia more often than child report [χ2 (1, N = 296) = 5.05, P < 0.05]. Physician evaluation was less likely to yield a “no diagnosis” classification (0%) than report by child (16%; P < 0.001, Fisher exact test) or parent (11%, P < 0.001, Fisher exact test). Physician evaluation also was less likely to yield a classification of IBS alone than child report [χ2 (1, N = 296) = 16.00, P < 0.001] or parent report [χ2 (1, N = 296) = 4.27, P < 0.05]. Physician evaluation was more likely to yield a combined IBS/FD classification than child report [χ2 (1, N = 296) = 19.30, P < 0.001] or parent report [χ2 (1, N = 296) = 19.30, P < 0.001].
The overall rates of agreement, kappa statistics and weighted kappa statistics are shown in Table 2. Parent report was more consistent with a child report diagnosis of dyspepsia than child-reported IBS [χ2 (1, N = 124) = 15.54, P < 0.001]. Kappa statistics indicated fair diagnostic agreement across all dyads (6). Weighted kappas were computed using only diagnoses of FD, IBS and combined IBS/FD to allow credit for partial agreement (i.e., IBS/FD with FD alone or IBS alone). Despite this increased flexibility with regard to diagnostic agreement, weighted kappa indicated fair-to-moderate diagnostic agreement across all dyads (6).
Reasons for Diagnostic Discrepancies
The reasons for discrepant diagnoses are summarized in Figure 2. The number and/or frequency of symptoms related to defecation was more likely to account for discrepant diagnoses than all other causes combined for each dyad [child versus physician: χ2 (1, N = 178) = 34.18, P < 0.001; child versus parent: χ2 (1, N = 172) = 37.21, P < 0.001; parent versus physician: χ2 (1, N = 164) = 31.61, P < 0.001].
Eighty-eight patients who had a diagnosis of FD or IBS/FD based on physician evaluation underwent upper endoscopy with biopsy. All patients had failed acid reduction therapy before biopsy. All had visually normal endoscopies with no erosions, ulcers or nodularity seen. Histologic evaluation of biopsies revealed the following: mild esophagitis (<4 eosinophils/high power field) in 21%; gastritis in 44%, including Helicobacter pylori-associated gastritis in 1 patient and chronic gastritis (mild-to-moderate increase in lymphocytes and/or plasma cells) in 38; patients and chronic duodenitis (mild to moderate increase in lymphocytes and/or plasma cells without villous blunting) in 25%. No patients had evidence of intense or severe inflammation.
We found that the majority of children referred to a tertiary center for evaluation of recurrent abdominal pain fulfill the Rome II criteria for FGID. This is true whether the diagnosis is made by child report, by parent report or by physician evaluation and judgment. These findings are consistent with Walker et al., who found that 73% of patients referred for RAP fulfilled criteria for an FGID (4).
Only 16% of patients did not meet Rome II criteria for an FGID by child report. The most common reason for not meeting criteria by child report was symptom duration of less than 12 weeks. Most children would have otherwise met Rome II criteria for an FGID. Our study did not specifically examine whether the 12-week time requirement for an FGID diagnosis is optimal. This time frame was originally proposed by Apley and Naish in the 1950s and has continued to be the standard for both adult and pediatric populations (2). Although we agree that a time requirement is needed to exclude children with acute illness (e.g., viral infection), it remains unclear whether 12 weeks represents the optimal cutoff point. At present, there are no data available to demonstrate that children with abdominal pain at 12 weeks differ substantially from children seen at 11 weeks, 8 weeks or even 4 weeks. We recommend that this time requirement be critically examined in future research to ensure its appropriateness.
Even if we accept the 12-week time requirement, our data suggest that some children are unreliable with respect to reporting on the timeline for their symptoms. For all of the children who reported fewer than 12 weeks of symptoms, parents reported a symptom duration greater than 12 weeks. In addition, most of these children also had tests, physician visits and/or treatments to address abdominal pain issues documented in their medical history that were consistent with a longer symptom course. With the time factor removed, over 93% of children in our study would have met Rome II criteria by self report. Thus, we recommend that children who meet symptom criteria should not be disqualified from receiving an FGID diagnosis based on self-reported symptom duration.
By parent report, 11% of patients did not meet Rome II criteria for an FGID. Failure to meet criteria generally related to a single associated symptom, such as relation of the pain to eating, menses or defecation. An additional 73% of children classified as “no diagnosis” by parent report would have met Rome II criteria if the single associated symptom was removed. It remains unclear whether these single symptoms distinguish these patients from other patients meeting FGID diagnostic criteria in a clinically meaningful way.
Walker et al. reported a systematic evaluation of Rome II criteria for abdominal pain related FGIDs in children/adolescents (4). Although we used the same questionnaire and scoring criteria, we found FD to be the most common diagnosis regardless of source, whereas IBS was the most common diagnosis found by Walker et al. Differences between the studies may result from referral patterns in the clinics from which the samples were drawn. For instance, children with a primary complaint of constipation are not typically sent to the Abdominal Pain Clinic, which was the primary data collection site for the current study (CMH). Instead, these children are referred to other gastrointestinal staff within the hospital system for treatment. Thus, the number of children with IBS symptoms in our study may be lower relative to other clinics to which children with constipation are referred.
Despite this referral bias, the FGID frequencies found in our study are consistent with other reports. Hyams et al. reported dyspepsia in 49% of patients referred for abdominal pain evaluation (7). Twenty-four percent of dyspeptic patients also were found to have IBS. In another study, Hyams et al. reported IBS symptoms in 68% of patients referred for abdominal pain (8). However, many of these patients also had dyspeptic symptoms, including epigastric pain in 31%, nausea in 48% and vomiting in 24%. Shaffer et al. reported dyspepsia in 73% of a referred group of children with abdominal pain (9). These results suggest that many children experience a combination of upper and lower symptoms consistent with FD and IBS and that diagnosis may vary depending on the perspective of the evaluating physician. Studies completed before Rome II criteria also may be based on differing diagnostic criteria.
Recent evidence suggests that evaluating physicians may vary widely in their understanding and application of the Rome II criteria. Pediatric gastroenterologists provided with a copy of the Rome II criteria show low inter-rater reliability for diagnosis when faced with clinical vignettes (10). These findings suggest that the Rome II criteria may need further refinement and clarification or, alternately, that further training of practitioners in the use of the current Rome II criteria is needed to reduce physician variability and standardize diagnosis.
Our results suggest a few areas in which the Rome II criteria may benefit from further refinement and clarification. We found that the child's defecation history and its interpretation were likely to lead to discrepant diagnoses. Differences in parent-child diagnoses often resulted from disparate estimates regarding the frequency of the child's stool symptoms. The QPGS requires a stool symptom frequency of “sometimes” or more for IBS and of “sometimes” or less for FD or FAP; these criteria are clearly written in the scoring procedures for this questionnaire and represent one interpretation of the Rome II criteria. Thus, even relatively minor differences in reports of symptom frequency could have a negative impact on parent-child diagnostic agreement. Similarly, differences in physician-parent and physician-child diagnoses often resulted from the physician judging that a particular stool symptom was of sufficient frequency and intensity to be meaningful, whereas the QPGS scoring criteria yielded the opposite outcome. The Rome II criteria allow latitude for physician judgment and are less specific about the frequency with which stool symptoms must occur to meet specific diagnoses than is the QPGS. Thus, the same diagnostic label may not be applied uniformly from physician to physician unless the Rome II criteria are made more specific.
Similarly, although diagnosis of an FGID requires that there be no evidence of organic disease to explain symptoms, the definition of organic disease remains problematic when applying Rome II criteria (3). Differences in diagnosis may arise when histologic inflammation is identified by the pathologist in the setting of a visually normal endoscopy. Although mucosal inflammation has been reported in up to 93% of children undergoing endoscopy for RAP (11), other reports indicate that these nonspecific histologic findings are subject to significant interobserver variance (12,13). Many of the most common forms of histologic inflammation have no clinical correlation or clear therapeutic implication (14). Finally, emerging adult data suggest that histologic inflammation actually may be involved in the pathogenesis of FGIDs such as IBS (15,16). We felt, on the basis of these studies, that histologic inflammation should not categorize a child as having organic disease but rather that these children should be grouped with those having FGIDs. The exception to this would be children found to have intense or severe inflammation, such as that associated with Crohn's disease, giardiasis, eosinophilic esophagitis and eosinophilic gastroenteritis diagnosed histologically (>40 eosinophils/high power field), as the significance of the inflammation in these situations is generally accepted as causative of symptoms and would result in classification as an organic disease. Clear guidelines on the interpretation of mild-to-moderate histologic inflammation are not provided in the Rome II criteria, so variance may exist between physicians that may impact clinical and research efforts. Future revisions of the Rome criteria should more clearly address this issue.
As long as the Rome criteria remain open to interpretation in these areas, inter-rater reliability among physicians is likely to be poor. As such, it is difficult to argue that physician evaluation should be considered the current diagnostic standard. More importantly, even if the Rome criteria are clarified with the next revision and inter-rater reliability improves, it is important to consider whether this standard would be clinically meaningful. To date, little research has been done to examine the benefit of categorizing FGIDs in this way. Research examining whether Rome diagnostic categories predict response to specific treatments is needed. This information could provide the empiric basis for additional refinement of the Rome II criteria.
Further investigation of the relationship between histologic inflammation and FGIDs also is warranted because of the potential for inflammation to alter gastrointestinal sensory and motor function. Improved understanding of this relationship could contribute to further Rome II refinement and perhaps to improved treatment for children with recurrent abdominal pain.
1. Oberlander TF, Rappaport LA. Recurrent abdominal pain during childhood. Pediatr Rev
2. Apley J, Naish N. Recurrent abdominal pains: a field study of 1000 school children. Arch Dis Child
3. Rasquin-Weber A, Hyman PE, Cucchiara S, et al. Childhood functional gastrointestinal disorders. Gut
4. Walker LS, Lipani TA, Greene JW, et al. Recurrent abdominal pain: symptom subtypes based on Rome II criteria for pediatric functional gastrointestinal disorders. J Pediatr Gastroenterol Nutr
5. Walker LS, Caplan-Dover A, Rasquin-Weber A. Manual for the Questionnaire on Pediatric Gastrointestinal Disorders. Nashville, TN: Department of Pediatrics
, Vanderbilt University School of Medicine, 2000.
6. Altman DG. Practical statistics for medical research. London: Chapman and Hall;1991.
7. Hyams JS, Davis P, Sylvester FA, et al. Dyspepsia in children and adolescents: a prospective study. J Pediatr Gastroenterol Nutr
8. Hyams JS, Treem WR, Justinich CJ, et al. Characterization of symptoms in children with recurrent abdominal pain: resemblance to irritable bowel syndrome. J Pediatr Gastroenterol Nutr
9. Shaffer SE, Sellman SB, Repucci AH, et al. Dyspepsia: redefining chronic abdominal pain in children. Gastroenterol
10. Saps M, Di Lorenzo C. Inter-rater reliability of the Rome criteria in children. J Pediatr Gastrointest Nutr
11. Mavromichalis I, Zaramboukas T, Richman PI, Slavin G. Recurrent abdominal pain of gastro-intestinal origin. Eur J Pediatr
12. Black DD, Haggitt RC, Whitington PF. Gastroduodenal endoscopic-histologic correlation in pediatric patients. J Pediatr Gastroenterol Nutr
13. Ashorn M, Ruuska T, Karikoski R, Valipakka J, Maki M. Gastric mucosal cell densities in Helicobacter Pylori
-positive and negative dyspeptic children and healthy controls. J Pediatr Gastroenterol Nutr
14. Mayer EA, Collins SM. Evolving pathophysiologic models of functional gastrointestinal disorders. Gastroenterol
15. Chadwick VS, Chen W, Shu D, et al. Activation of the mucosal immune system in irritable bowel syndrome. Gastroenerol
16. Drossman DA, Camilleri M, Mayer EA, Whitehead WE. AGA technical review on irritable bowel syndrome. Gastroenterol