Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
Constipation is a common problem in children. Diagnosis is based on clinical features. In case of doubt about the presence of constipation, the existence of fecal retention can be evaluated on plain abdominal radiography.
To describe and to assess the evidence from observational, controlled studies concerning the association between abdominal radiography and symptoms and signs related to constipation in children.
MEDLINE was searched from inception to April 2004 using a specified search strategy. Studies that fulfilled predefined criteria were assessed for methodological quality. Study characteristics and associations were extracted and the results were summarized according to a best-evidence synthesis.
Of the 392 publications identified, 6 studies met the inclusion criteria. Only 2 studies were of high methodological quality. The best-evidence synthesis yielded conflicting evidence for an association between a clinical and a radiological diagnosis of constipation. The likelihood ratio (LR) in 2 high-quality studies was close to 1 (LR, 1.2; 95% confidence interval [CI], 1.0-1.4; and LR, 1.0; 95% CI, 0.5-1.6). Conflicting evidence was found for an association between digital rectal examination and fecal impaction on radiography. Limited evidence was found for an association between a history of hard stool and a finding of rebound tenderness and radiography (LR, 1.2; 95% CI, 1.0-1.4; and LR, 1.1; 95% CI, 1.0-1.2, respectively).
The limited amount of data available shows conflicting evidence for an association between clinical symptoms of constipation and fecal loading on abdominal radiographs in children. The recommendation to perform a plain abdominal radiograph in case of doubt of the presence of constipation in a child cannot be supported. Further research of good methodological quality is needed.
Constipation with or without encopresis is a common problem in children. It affects 3% of preschool-aged children and 1% to 2% of school-aged children.1 Peak incidence occurs at the time of toilet training, between the ages of 2 and 4 years.2Defecation disorders represent the chief reports in 3% of pediatric outpatient visits and 10% to 25% of pediatric gastroenterology visits.3 The general well-being of adult patients with chronic constipation is lower than that of a comparable normal population.4 In children it is well known that encopresis sometimes leads to social withdrawal, low self-esteem, and even depression.5 Longer duration of symptoms before diagnosis has been associated with poorer long-term outcome.2
In more than 90% of the children with constipation and encopresis, no organic or anatomical cause can be found and, therefore, these patients are considered as having a functional defecation disorder. These patients have been later classified by the Paediatric Rome Criteria as having functional constipation, functional fecal retention, or functional nonretentive fecal soiling.3
The key features of pediatric constipation are a defecation frequency fewer than 3 times per week in combination with the involuntary loss of stool in the underwear.6 Other important clinical parameters are stool consistency, large-caliber stools, pain at defecation, and abdominal discomfort.7 However, there are no well-designed studies that determine which aspects of a medical history and physical examination are most important in discriminating between constipation and nonconstipation.8
In case of doubt about the presence of constipation one would like to evaluate the existence of fecal retention or impaction. More than a decade ago Barr et al9 introduced a score to appraise fecal retention on a single radiograph of the abdomen. Since then different scoring systems have been developed to assess fecal loading on an abdominal radiograph.10,11 However, interobserver and intraobserver agreement on the existence of constipation on abdominal radiography varies between the scoring systems used.
To evaluate the additional diagnostic value of abdominal radiography in the diagnosis of constipation in children, we performed a systematic literature review. The objective was to describe and to assess the evidence from observational, controlled studies concerning the association between abdominal radiography and symptoms and signs related to constipation in children.
Is a score of fecal loading on abdominal radiography a sensitive and specific test compared with a diagnosis of constipation based on symptoms and signs in children?
MEDLINE was searched from inception to April 2004 using the terms “fecal impaction,” “coprostasi,” “encopresis,” “constipation,” and “obstipation” in combination with “radiography, abdominal,” “medical history taking,” and/or “physical examination.” All terms were included as MeSH heading and as text word. The results of this search were combined with the search strategy for identifying diagnostic studies, as described by Devillé and Buntinx.12
Additional strategies for identifying trials included searching the reference lists of review articles and included studies. Experts in the field were asked to identify further published and unpublished primary studies.
Only controlled, observational studies investigating the relation between fecal loading on plain abdominal radiography and symptoms and signs related to constipation in otherwise healthy children aged from 1 to 18 years were eligible for inclusion in the review. The reference tests considered included fecal loading on plain abdominal radiography according to a predefined scoring system, as well as a clinical diagnosis of constipation according to the presence or absence of predefined symptoms and signs. Results of the study had to allow for the extraction of information on diagnostic value (sensitivity, specificity, likelihood ratio [LR], or accuracy). No language restriction was applied.
Two reviewers (L.M.R.-V. and M.Y.B.) independently screened titles and abstracts of studies identified by the literature search for eligibility. All potentially relevant studies were retrieved as full papers and then again independently reviewed by 2 reviewers (L.M.R.-V. and M.Y.B.). Decisions regarding the inclusion of studies were made independently and any disagreements were resolved through consensus whenever possible, or by arbitration of a third reviewer (S.B.-Z.).
Two reviewers (L.M.R.-V. and M.Y.B. or S.B.-Z.) independently assessed the methodological quality of the included studies using the QUADAS instrument13 (Table 1). All selected methodological criteria were scored as yes (1), no (0), or do not know (0). The overall methodological quality of a study was computed by counting the number of positive scores, with equal weights applied on all items. Scores could range from 0 to 14. We arbitrarily regarded trials with methodological quality scores of 9 or higher (>60%) as being of high quality. In case of a disagreement between the 2 reviewers, consensus was used to resolve disagreement. When consensus could not be reached, a third reviewer made the final decision (M.Y.B. or S.B.-Z.).
Two reviewers (L.M.R.-V. and M.Y.B.) performed independently a structured extraction of data from the original reports. Disagreements were resolved by consensus. Extracted information included (if available) demographic data, a definition of the participants with and without the disease as determined by the reference test, detailed description of the reference test and index test, reproducibility of the index test, and outcome measures.
The interassessor agreement on the methodological quality was calculated using κ scores. The κ values range from −1, indicating perfect disagreement, to +1, indicating perfect agreement. A κ value greater than 0.7 indicates a high level of agreement between assessors; a κ value between 0.5 and 0.7, a moderate level of agreement; and a κ value less than 0.5, a poor level of agreement.14 Wherever possible we calculated sensitivities, specificities, and LRs with a 95% confidence interval (CI).
In case of clinical heterogeneity (patient population and/or definition of reference and index test are not considered to be sufficiently similar), the results were not pooled but a best-evidence synthesis was used to summarize these data. The level of evidence was then ranked. Strong evidence exists when multiple high-quality studies show consistent findings, moderate evidence exists when one high-quality study and multiple low-quality studies show consistent findings, and limited evidence exists when one high-quality study is found or multiple low-quality studies show consistent findings. Findings are considered consistent when 75% or more of the studies reported the same result. Only statistically significant associations are considered as associated in this synthesis. A statistically significant association exists when the 95% CI of the LR does not include 1.
A total of 392 publications were identified by our search strategy. Of these, 12 studies met our initial inclusion criteria.9- 11,15- 23 Seven studies were excluded for the following reasons: in the study of Blethyn et al19 only children with urinary tract infections were included, van der Plas et al22 did not include a control group, in the study of Bewley et al18 no data on diagnostic value were presented, and in 4 studies the symptoms of constipation were not related to the outcome of a plain abdominal radiography.15,19,20,23 No additional studies could be retrieved from the reference lists of relevant articles. Therefore, a total of 6 studies were included in this review.9- 11,16,17,21 Full details of the included studies are presented in Table 2 and Table 3.
We included 3 case series,16,17,21 2 case-control studies9,11 and 1 study of a retrospective reexamination of abdominal radiographs.10 In 1 case-series17 the association between the radiographic diagnosis of constipation and the clinical diagnosis of constipation was evaluated in a selection of all included children. In our analysis this study was further considered as a case-control study. All studies were hospital based.
In only 1 study was the purpose to determine whether clinical variables accurately identify children with radiographically proven constipation (Table 3).16 In the 6 studies included, 3 different scoring systems for assessing fecal impaction on abdominal radiography were used. Three studies used the Barr-score,9,17,21 2 studies used a revision of the Barr-score,16,19 and Leech et al11 rated according to their own scoring system.
In 3 studies a clinical diagnosis of constipation was considered as the reference test.9,11,17 Only 2 studies gave a good description of the diagnosis of constipation9,17 (Table 3). In the study of Benninga et al17 patients with soiling or encopresis were eligible when they had at least one other symptom of constipation, whereas in the study of Barr et al9 children with soiling or encopresis were excluded. In the 2 case series the abdominal radiograph was considered as the reference test16,21; 1 of these 2 studies retrospectively retrieved its information on symptoms and signs of constipation from medical record forms.21 In the latter study only children with encopresis were included.
The reviewers scored 84 items and agreed on 65 items (77.4%, κ = 0.54) (Table 1). The 19 disagreements were resolved in a single consensus meeting. Disagreement was largest at item 4 (50%): “Is the period between reference standard and index test short enough to be reasonably sure that the target condition did not change between the 2 tests?” The total scores ranged from 5 to 11. Items 4, 12, and 13 were the least fulfilled criteria (1, 2, and 0 times, respectively). The mean methodological quality score was 8.3.
Beckmann et al16 performed a multivariate discriminant analysis. In a logistic regression model, clinical symptoms of constipation were introduced as the independent variables and radiographic constipation as the dependent variable. The model defined a clinical diagnosis of constipation. This clinical diagnosis existed 1.2 times as often in children with a radiographic diagnosis of constipation as in children without a radiographic diagnosis of constipation (LR, 1.2; 95% CI, 1.0-1.4) (Table 4).
Four studies examined whether abdominal radiography was able to discriminate between children with clinical constipation and children without clinical constipation.9- 11,17 Only Leech et al11 found a statistically significant discriminative value (LR, 3.0; 95% CI, 1.6-4.3) (Table 4). Blethyn et al10 presented accuracy as a measure for the ability of abdominal radiography to discriminate between clinically constipated and nonconstipated children (accuracy, 80%; 95% CI, 50%-100%).
Diagnostic value was related to the methodological quality of the studies. Low-quality studies9,11 reported better diagnostic values compared with high-quality studies16,17 (Figure). In particular the specificity was poor in the high-quality studies (Table 4).
The reported diagnostic association between clinical symptoms and radiographic obstipation. Numbers are quality scores. the x- and y-axes present proportions. In case the sensitivity equals 1-specificity and the likelihood ratio of the test equals 1. That means the test does not discriminate between constipated and nonconstipated children.
The diagnostic value of individual symptoms was evaluated in 2 studies.16,21 Results are presented in Table 5. Beckmann et al16 defined a positive rectal examination as stool present in the rectal vault. They found a significant LR for a positive rectal examination (LR, 1.6; 95% CI, 1.2-2.0). However, Rockney et al21 did not confirm this finding (LR, 1.5; 95% CI, 0.8-2.3). They defined a positive rectal examination as the presence of moderate or large amounts of stool on rectal examination.
Rockney et al21 found a significant LR for a history of hard stool (LR, 1.2; 95% CI, 1.0-1.4). Beckmann et al16 found a significant LR for a finding of rebound tenderness (LR, 1.1; 95% CI, 1.0-1.2).
Only in the study of Benninga et al17 did interobserver reliability of the rating of the radiographs range from poor to moderate (κ = 0.28-0.60). All other studies reported moderate to excellent interobserver reliability (κ range, 0.6321-0.9510) (Table 3). Intraobserver reliability was evaluated in 3 studies9,11,21 and ranged from moderate (κ = 0.5221) to excellent (κ ≥ 0.859) (Table 3).
One high-quality study16 and 1 low-quality study11 found a statistically significant association between a clinical diagnosis of constipation and a radiographic diagnosis of constipation (50% of the findings). One high-quality study17 and 1 low-quality study9 did not find a statistically significant association (50% of the findings). These findings are inconsistent (conflicting evidence). One high-quality study found a significant association between a finding of stool at rectal examination and radiographic finding of constipation.16 This association was not statistically significant in another low-quality study (conflicting evidence).21 One high-quality study found a significant association between rebound tenderness and radiographic constipation (limited evidence).16
This systematic review of the literature was performed to investigate the possible association between clinical symptoms and radiographic features of childhood constipation. We identified a total of 6 studies, that were eligible for inclusion in this study. Only 2 of these studies were of good methodological quality.16,17
We found conflicting evidence for an association between a clinical and a radiographic diagnosis of constipation. Methodological quality of the study was related to the diagnostic value reported by the studies. High-quality studies (score ≥9) found LRs close to 1 in contrast to low-quality studies that reported high LRs (Figure).
The results of rectal examination were not consistently related to findings on fecal retention at abdominal radiography. Likelihood ratios of a finding of stool at rectal examination were close to 1, indicating that stool at rectal examination occurred almost as often in children with fecal loading on radiography as in children without fecal loading (Table 5). The time between rectal examination and abdominal radiography might have influenced the accuracy of the abdominal radiography. In the high-quality study abdominal radiography was performed almost directly after rectal examination.16 In the low-quality study this period was not specified.21
The interreliability and intrareliability of scoring systems for fecal loading on abdominal radiography varied among studies from poor to excellent. Experienced radiologists had more reliable scores than less experienced colleagues. Benninga et al17 investigated the reproducibility of the Barr-score on 2 consecutive radiographs 3 days apart, without administering a laxative. The interobserver reliability between the 2 radiographs on the existence of constipation as measured by a Barr-score of 10 was poor. In their study the radiographic diagnosis of constipation in a child was not reproducible.
Constipation can be diagnosed by a detailed medical history and a thorough physical examination including a digital rectal examination. Symptoms of constipation like stool frequency, encopretic episodes, and the passage of large amounts of stools are reliably recalled by the child or the child’s parents.24 Nevertheless a diagnosis might be doubtful in case not all key symptoms of constipation are present or when a rectal examination is infeasible. The clinical practice guideline of the North American Society for Pediatric Gastroenterology and Nutrition for diagnosing constipation in children recommends taking a plain abdominal radiograph in case of doubt about the presence of constipation, in case the child refuses a rectal examination, or if rectal examination is considered traumatic.8 The results of our review do not support this recommendation. We found conflicting evidence for an association between a clinical diagnosis of constipation and a radiographic diagnosis. Furthermore, high-quality studies found LRs close to 1 indicating that a radiographic diagnosis of constipation occurs almost as often in clinically constipated as in clinically nonconstipated children.
Although there was wide heterogeneity between the 6 studies, all studies were homogeneous in their hospital-based setting. Therefore, the results of our review cannot be generalized to general practice.
Although the literature search was extensive and had no language restriction, it might have a risk of publication bias. We found 6 studies of which only 2 were of good methodological quality. Therefore, one missed high-quality study may easily alter the results of our best-evidence synthesis. However, because none of the consulted experts in the field could add any missed or ongoing study, it is unlikely that such a study exists.
In this review only statistically significant LRs were considered of diagnostic value. This criterion may misclassify the results of studies with a small sample size (introduction of type II error due to the low statistical power of small studies). The conflicting evidence we found for the association between a clinical and a radiographic diagnosis of constipation was not influenced by the sample size of the studies. In contrast, small sample size may be of influence in our finding of conflicting evidence for an association between stool on rectal examination and radiographically confirmed constipation. In a study of 251 children Beckmann et al16 reported a statistically significant LR of 1.6 (95% CI, 1.2-2.0); Rockney et al21 studied 60 children and reported an LR of 1.5 but did not reach statistical significance (95% CI, 0.8-2.3). Although pooling of data would have been a solution to overcome the problem of small sample size, we refrained from pooling because of the substantial differences between studies.
Most clinicians consider a clinical diagnosis of constipation as the gold standard. However, there are no well-designed studies that determine which aspects of a medical history and physical examination are most important in discriminating between constipation and nonconstipation.8 de Lorijn et al25 found a good relation between clinical symptoms and colonic transit time. However, 50% of their patients who fulfilled the clinical criteria for constipation had colonic transit times within the normal range. Therefore, one might argue that a clinical diagnosis of constipation as well as fecal loading on radiographs are both substitutes of an adequate reference standard for constipation in children. In case of lack of an adequate reference standard, follow-up studies (preferably randomized) to quantify the effect of a diagnostic test on patient outcome are needed. Evaluating a test on patient outcome involves the evaluation of the diagnostic tests (clinical diagnosis, abdominal radiography, and colonic transit time) plus current administered therapies (laxatives and behavioral interventions) combined.26 Given the LRs of almost 1 in the high-quality studies, one should discuss whether the diagnostic strategy evaluated in such a study should incorporate plain abdominal radiography.
In the limited data available we found conflicting evidence for a diagnostic association between clinical symptoms of constipation and fecal loading in abdominal radiographs in children. The recommendation to perform a plain abdominal radiograph in case of doubt of the presence of constipation in a child cannot be supported by this systematic review. Further well-powered research of good methodological quality is needed to find the best diagnostic strategy in children suspected of having constipation.
Correspondence: M. Y. Berger, Department of General Practice, Erasmus Medical Centre, K Ff337, PO Box 1738, 3000DR Rotterdam, the Netherlands (email@example.com).
Accepted for Publication: February 13, 2005.
Reuchlin-Vroklage LM, Bierma-Zeinstra S, Benninga MA, Berger MY. Diagnostic Value of Abdominal Radiography in Constipated ChildrenA Systematic Review. Arch Pediatr Adolesc Med. 2005;159(7):671-678. doi:10.1001/archpedi.159.7.671