Abdominal Radiography Findings in Small-Bowel Obstruction: Relevance to Triage for Additional Diagnostic Imaging

John C. Lappas 1 Benedicto L. Reyes 2 Dean D. T. Maglinte 3

OBJECTIVE. Our aim was to determine which ndings on abdominal radiography are relevant for distinguishing complete or high-grade partial small-bowel obstruction from lowgrade partial or no small-bowel obstruction. MATERIALS AND METHODS. Admitting abdominal radiographs with the patients in the supine and upright positions were scored for 25 different ndings in 81 patients with clinically suspected small-bowel obstruction. Forty-one patients had complete or high-grade partial small-bowel obstruction, and 40 had low-grade partial small-bowel obstruction or no obstruction as determined by enteroclysis examination. Abdominal radiography ndings were subjected to statistical analysis for correlation with degree of obstruction. RESULTS. Of 12 radiographic ndings strongly associated ( p < 0.05) with the severity of obstruction, two ndings were found to be the most signicant ( p 0.0003) and predictive of a higher grade small-bowel obstruction: the presence of airuid levels of differential height in the same small-bowel loop and the presence of a mean airuid level width greater than or equal to 25 mm on upright abdominal radiographs. CONCLUSION. When both critical ndings are present, the degree of small-bowel obstruction is likely high-grade or complete. When both signs are absent, small-bowel obstruction is likely low-grade or nonexistent. Upright abdominal radiographs are important in the examination of patients with suspected small-bowel obstruction and may contribute to the imaging triage of these patients.

Received February 29, 2000; accepted after revision June 12, 2000.
1

Department of Radiology, Indiana University School of Medicine, Wishard Memorial Hospital, 1001 W. Tenth St., Indianapolis, IN 46202. Address correspondence to J. C. Lappas.

Department of Radiology, St. Francis Hospital, 1500 Albany St., Beech Grove, IN 46107.

3 Department of Radiology, Indiana University School of Medicine, 550 N. University Blvd., Indianapolis, IN 46202.

AJR 2001;176:167174
0361803X/01/1761167 American Roentgen Ray Society

ccording to a recent report, conventional abdominal radiography remains the preferred method of initial radiologic examination of symptomatic patients suspected of smallbowel obstruction [1]. These investigators and others advocate that if further diagnostic imaging is warranted to establish a more specic diagnosis or to obtain information pertinent to the clinical treatment, the results of the abdominal radiographic examination should serve as the basis for triage [1, 2]. It is recommended that if the initial abdominal radiographs suggest a complete or highgrade partial small-bowel obstruction, CT would be the preferred additional imaging modality if surgery is not imminently planned. This approach is supported by the sensitivity of CT for establishing the diagnosis of a high-grade or complete small-bowel obstruction, reported to be 82100%, and the potential value of CT to modify the patients treatment [1, 37]. Conversely, if the initial

radiographic ndings are interpreted as normal, equivocal, or suggestive of a low-grade partial small-bowel obstruction, an examination using a direct contrast material infusion of the intestinal lumen is preferred. Enteroclysis and CT enteroclysis satisfy these requirements and are recommended on the basis of their high sensitivity for the diagnosis of a less severe or low-grade small-bowel obstruction, their ability to exclude the diagnosis of obstruction, and the comparatively lower diagnostic sensitivity of conventional CT in this clinical situation [810]. Experienced radiologists have shown accuracy and agreement for the diagnosis of acute high-grade small-bowel obstruction on the interpretation of abdominal radiographs, but the diagnosis of low-grade obstruction is less certain [8, 11]. Additionally, no substantial data exist regarding the capability of abdominal radiography to distinguish the severity of obstruction of the small bowel. In a study of patients with a clinical suspicion of small-

AJR:176, January 2001

167

Lappas et al. bowel obstruction who were later proven to have various degrees of small-bowel obstruction or no obstruction, we objectively determined the frequency and pertinent measurement of the various ndings on abdominal radiographs that are commonly used to evaluate small-bowel obstruction. Considering the implications for subsequent patient workup based on the use of abdominal radiographs as a diagnostic triage tool, we attempted to establish which ndings correlated with the degree of smallbowel obstruction and to determine from those ndings which were the most signicant in differentiating the severity of smallbowel obstruction.
Materials and Methods
All patients were recruited retrospectively by review of the radiology clinical database according to the study requirements as follows: a diagnosis of small-bowel obstruction was suspected clinically and was the primary reason for referral for imaging, admitting abdominal radiographs were obtained with the patient in the supine and upright positions and were available for review, and an enteroclysis examination was performed within 5 days of the admitting abdominal radiography. Study patients were enrolled in a chronologically consecutive fashion until an equivalent number of patients constituted each of four groups: 21 patients had complete small-bowel obstruction, 20 had high-grade and 20 had low-grade partial small-bowel obstruction, and 20 had no obstruction. Thus, a total of 81 patients constituted the study population. Findings at enteroclysis examination served as the basis for the assignment of patients to a study group. Enteroclysis examinations were performed by senior staff radiologists (including one of the authors) with expertise in gastrointestinal imaging or by residents directly supervised by the staff radiologist. All enteroclysis examinations were reviewed retrospectively and independently of the analysis of the abdominal radiographs and were evaluated by consensus of two senior gastrointestinal radiologists to accurately determine the grade of small-bowel obstruction or the nding of no obstruction. The abdominal radiograph routinely obtained as a preliminary scout view with the enteroclysis examination was deleted from the enteroclysis review. For the diagnosis of low-grade, high-grade, and complete small-bowel obstruction and for the exclusion of mechanical obstruction on the enteroclysis examination, we followed the denitions in published criteria [1, 8, 9]. Obstruction was diagnosed when a transition zonedened as a change in caliber of the small-bowel lumen from a distended segment proximal to the site of obstruction to a segment that was decreased in caliber or collapsed distal to the obstructionwas seen at enteroclysis. Our enteroclysis criteria were 3.5 cm as the upper limit for normal caliber of the jejunal lumen and 2.5 cm as the upper limit for the ileal lumen. Low-grade partial small-bowel obstruction was diagnosed when the contrast media arrived at the transition zone without delay and sufcient contrast material owed through the obstruction into the loops beyond the obstruction so that mucosal fold patterns in these distal loops were clearly dened (Fig. 1A). High-grade partial small-bowel obstruction was diagnosed when there was a denite delay in the ow of contrast media to and through the site of obstruction and only small amounts of contrast material entered the collapsed bowel loops beyond the obstruction (Fig. 1B). Complete small-bowel obstruction was diagnosed by an absence of contrast media entering the bowel loops distal to a discrete point of lumen obstruction (Fig. 1C). The diagnosis of mechanical small-bowel obstruction was excluded at enteroclysis when unimpeded ow of contrast media was observed uoroscopically from the duodenojejunal junction to the colon and a normal-caliber small bowel without transition zones was documented. The patients nal discharge diagnoses, including pertinent clinical and laboratory data, surgical ndings, and diagnostic ndings on enteroclysis, were used to determine the cause of small-bowel obstruction and the cause of abdominal symptoms in the patients without small-bowel obstruction. Abdominal radiographs were reviewed retrospectively and in random order, and the evaluation was made by consensus of the authors, who included two senior gastrointestinal radiologists and a senior radiology resident. Evaluations were made without knowledge of the corresponding enteroclysis result or the patients clinical outcome. The patients supine and upright abdominal radiographs at admission were scored with respect to 25 different variables, 17 of which involved a specic numeric measurement (continuous variable) and eight of which involved the presence or absence of a condition or variable (categoric variable). A diagnostic interpretation of the abdominal radiographs was not performed as part of the scoring evaluation. The specic variables scored on the supine radiograph included the number of distended ( 2.5cm lumen diameter) or nondistended (< 2.5-cm lumen diameter) gas-lled or uid-lled smallbowel loops; the maximum and mean small bowel diameter (in millimeters); and the maximum and mean diameter (in millimeters) of various colon segments, including the cecum, ascending colon, transverse colon, descending colon, and rectum (Fig. 2). Presence of the stretch sign, dened as small-bowel gas arrayed as stripes perpendicular to the long axis of the bowel, as an indicator of a predominantly uid-lled small-bowel loop, was noted (Fig. 3). Comparative ratios of small-bowel size to colon size were calculated from the maximum lumen diameter measurements, and the occurrence of small bowelcolon ratios greater than 0.5 and greater than 1.0 were recorded. Variables scored on the upright radiograph included the number and width (in millimeters) of airuid levels measuring greater than or equal to 10 mm in width; the presence and differential height (in millimeters) of airuid levels in the same small-bowel loop (Fig. 4); the presence of the string-of-beads sign, dened as a series of air uid levels individually measuring less than 10 mm in width (Fig. 4); and the presence of a distended stomach. Statistical Analysis Two categories of small-bowel obstruction were dened for statistical analysis of the data. A severe or higher grade group of obstruction (n = 41) included patients with complete and highgrade partial small-bowel obstruction. A less severe or lower grade group (n = 40) included patients with low-grade partial small-bowel obstruction and patients with no obstruction. In the initial screening process, each scored radiographic nding or variable was examined on a univariate basis to evaluate its individual importance. During this screening, Fishers exact test was used for the categoric variables. For the continuous (numeric) variables, t tests were performed. All variables, categoric or continuous, yielding univariate p values of less than 0.3 were considered relevant and potentially signicant in determining the severity of small-bowel obstruction. To transform signicant continuous (numeric) variables into more clinically applicable radiographic ndings, cut-points (threshold values) were generated using tree-based modeling [12]. This method identied the optimal threshold value for continuous variables that would yield the best possible statistical separation between the categories of higher grade and lower grade small-bowel obstruction. Variables transformed in this manner would then be more appropriate for use in predicting the severity of small-bowel obstruction. Finally, a backward elimination logistic regression model was used to eliminate relevant variables that contained similar or noncontributory information [13, 14]. All explanatory variables ( p < 0.3) were initially included in the model, and nonsignicant variables were removed one at a time until only signicant variables remained. The process optimally generates a concise nal version of the predictive statistical model. All methods of statistical analysis were performed using S-Plus (StatSci, a division of MathSoft, Seattle, WA) and SAS (SAS Institute, Cary, NC) software.

Results

Enteroclysis examinations were performed within the rst 72 hr of admission in 34 (83%) of 41 patients in the higher grade category of obstruction and in 35 (88%) of 40 patients in the lower grade category of obstruction. Only 15% (12/81) of enteroclysis examinations were performed on either the

168

AJR:176, January 2001

Abdominal Radiography of Small-Bowel Obstruction

Fig. 1.Grades of small-bowel obstruction. A, Low-grade partial obstruction in 70-year-old man. Enteroclysis shows contrast medium in dilated small-bowel loop owing (curved arrow ) into discrete transition zone, represented by compressed lumen segment (straight arrow ). Contrast medium enters distal (D) bowel at rate of ow sufcient to achieve moderate lumen distension and to reveal mucosal folds. B, High-grade partial obstruction in 72-year-old woman. Enteroclysis shows contrast medium in dilated small-bowel loop owing (curved arrow ) into severely compressed lumen segment (straight arrows ). Rate of ow is impeded to degree that small amounts of contrast medium enter collapsed distal (D) bowel loops that have no dened mucosal folds. C, Complete obstruction in 84-year-old man. Enteroclysis shows that contrast medium fails to ow (curved arrow ) through completely compressed lumen segment (straight arrow ), and bowel loops distal to obstruction are not visualized.

C fourth or fth day after the patients admitting abdominal radiographs. Of the 41 patients in the higher grade category of obstruction, the cause of the obstruction was adhesions in 28 patients, neoplasm in six, ischemic stricture in three, hernia in two, intussusception in one, and perforated appendicitis in one patient. Of the 40 patients in the lower grade category of obstruction, low-grade partial obstruction was due to adhesions in 19 patients and Crohns disease in one patient, whereas patients without small-

AJR:176, January 2001

169

Lappas et al. bowel obstruction showed either nonobstructive adhesions (6 patients) or various conditions including enteritis in ve patients, chronic abdominal pain in four, gastroparesis in two, adynamic ileus associated with medication or electrolyte imbalance in two, and transient intussusception in one patient. The frequency of various (categoric) radiographic ndings in the two dened categories of small-bowel obstruction is summarized in Table 1. Table 2 summarizes the means and standard deviations of the numeric variables scored for each of the two categories. Table 3 lists the results of the initial screening process. Twelve variables were strongly associated with each group, including the maximum small-bowel diameter, the mean airuid level width, and the presence of a differential airuid level. All variables had p values of less than 0.05. One nding, the string-of-beads sign, was marginally associated with each group, which in this study was dened as a p value greater than 0.05 and equal to or less than 0.15. Twelve other variables, including the number of airuid levels, the measurements of colon size, the presence of colonic gas or feces, and any specic ratio of the maximum small-bowel diameter to maximum colon diameter, had no signicant correlation with the two groups ( p > 0.3). Table 4 lists the threshold values generated for each of the signicant continuous (numeric) variables. The p values listed are an approximate measurement of the statistical separation the threshold value creates between the two groups. Of the thirteen variables that passed the initial statistical screening process ( p < 0.3, Table 3), two were determined by logistic regression analysis to be the most signicant predictors between the two categories of severity of small-bowel obstruction. These were the presence of airuid levels of differential height in the same bowel loop and the presence of a mean airuid level width greater than or equal to 25 mm. Their strength as markers of a small-bowel obstruction of higher grade can be expressed as an odds ratio. The ratio is 7.2 for a mean air uid level width of 25 mm or greater, and the ratio is 4.5 for the presence of a differential airuid level. Table 5 presents the data for the association of the two most signicant radiographic ndings with the category of obstruction. The presence of both ndings meant an almost 86% chance that the patient had a high-grade partial or complete

Fig. 2.Abdominal radiograph with patient supine in 27year-old woman with high-grade partial small-bowel obstruction shows method of bowel lumen diameter measurement. Small-bowel loops (dotted lines ) and colon segments (solid lines ) were measured from widest bowel margin transverse to long axis of bowel. Left lower quadrant small-bowel loop measures 32 mm in diameter, and segment of transverse colon measures 40 mm in diameter.

Fig. 3.Abdominal radiograph with patient supine in 34-year-old man with complete small-bowel obstruction. Stretch sign is depicted by abnormal distention of predominantly uid-lled smallbowel loop where luminal gas has striped appearance oriented perpendicular to long axis of bowel (arrows ).

Fig. 4.Abdominal radiograph with patient upright in 36-year-old woman with high-grade partial small-bowel obstruction. Small-bowel loops A and B show differential height of airuid levels (dotted lines ) and were measured as vertical height (solid arrows ) between corresponding airuid levels in same bowel loop. Differential airuid height measures 31 mm in loop A and 5 mm in loop B. String-of-beads sign is depicted by airuid levels measuring less than 10 mm in diameter (open arrow ).

170

AJR:176, January 2001

Abdominal Radiography of Small-Bowel Obstruction

Frequency of Abdominal Radiography Findings According to Category of Small-Bowel Obstruction Higher Grade Finding No. Distended stomach Differential airfluid levels Stretch sign String-of-beads sign Presence of cecal fluid Presence of colon gas or feces Small bowelcolon ratio > 0.5 Small bowelcolon ratio > 1.0 7/41 31/41 14/41 16/41 6/41 36/41 36/41 16/41 % 17 76 34 39 15 88 88 39 No. 8/40 14/40 2/40 9/40 6/40 7/40 27/40 10/40 % 20 35 5 23 15 18 68 25 Lower Grade

TABLE 1

Note.Findings are categoric variables. Higher grade category ( n = 41) includes high-grade partial and complete obstruction, lower grade category (n = 40) includes low-grade partial obstruction and no obstruction.

TABLE 2

Mean and Standard Deviation (SD) Measurements of Abdominal Radiography Findings According to Category of Small-Bowel Obstruction Higher Grade Finding Mean SD 2.55 1.15 2.15 21.5 20.1 21.7 15.0 19.5 11.0 10.4 18.8 14.4 3.31 19.4 9.88 11.4 8.59 Mean 1.55 0.27 2.83 19.6 22.8 21.1 12.6 13.6 25.7 21.6 35.0 4.02 4.38 22.3 16.0 5.15 4.56 SD 1.91 0.51 2.60 21.0 16.4 21.0 12.2 19.1 10.3 8.49 17.2 10.1 3.87 19.2 11.3 8.64 7.94 Lower Grade

With patient supine No. of gas-filled small-bowel loops > 2.5 cm No. of fluid-filled small-bowel loops > 2.5 cm No. of gas- or fluid-filled small-bowel loops < 2.5 cm Cecum width (mm) Ascending colon width (mm) Transverse colon width (mm) Descending colon width (mm) Rectum width (mm) Maximum small-bowel diameter (mm) Mean small-bowel diameter (mm) Maximum colon diameter (mm) Ratio of small-bowel diameter to colon diameter With patient upright: No. of airfluid levels > 10 mm Maximum airfluid level width (mm) Mean airfluid level width (mm) Maximum differential airfluid level (mm) Mean differential airfluid level (mm)

3.37 0.98 1.39 19.3 19.0 16.7 11.7 12.0 35.7 29.0 36.9 6.13 5.39 37.0 25.6 10.6 8.80

Note.Findings are continuous (numeric) variables. Higher grade category ( n = 41) includes high-grade partial and complete obstruction, lower grade category (n = 40) includes low-grade partial obstruction and no obstruction.

small-bowel obstruction, because 18 of the 21 patients who showed the concurrent presence of both signicant ndings on abdominal radiographs were in the category of higher grade obstruction. When neither nding was present there was an 83% chance that the patient had a low-grade partial small-bowel obstruction or no obstruction, because 24 of the 29 patients who had neither nding on radiographs were in the lower grade category. Both or neither of these ndings was seen in 50 (62%) of 81 cases.

Overall, the presence of a differential air uid level on the upright abdominal radiograph was noted in 56% (45/81) of patients. A mean airuid level width greater than or equal to 25 mm was noted on the upright radiograph in 35% (28/81) of patients.

Discussion

Abdominal radiographs provide valuable cost-effective information in the early exami-

nation of patients with various abdominal symptoms and disorders, particularly intestinal obstruction. Limitations in the diagnosis of small-bowel obstruction are recognized, however, because the intestinal gas pattern upon which the interpretation of radiographs is largely based varies with certain characteristics such as the severity, level, and duration of the obstruction. In patients for whom abdominal radiographs are insufcient to condently establish a diagnosis of small-bowel obstruction or to reasonably assess the severity of obstruction, additional diagnostic imaging becomes necessary. The methodology in this study of combining symptomatic patients into two functional categories of obstruction was based on the appropriateness of the additional imaging evaluation as related to its sensitivity for the diagnosis of small-bowel obstruction and the implications for subsequent clinical management [17]. The grouping of patients with low-grade partial small-bowel obstruction with patients with no obstruction into a lower grade obstruction category is justied on the basis of their presenting symptoms being consistent with intestinal obstruction and on the basis that the recommended method of additional radiologic workup is similar for these patients. Although patients with low-grade partial small-bowel obstruction usually present with abnormal abdominal radiographic ndings diagnostic of obstruction, approximately 40% of such patients have normal or equivocal ndings [1, 8]. Similarly, even symptomatic patients who are eventually shown to have no small-bowel obstruction may have equivocal radiographic ndings suggestive enough to warrant further imaging to exclude obstruction. The sensitivity and accuracy of enteroclysis for dening a minimal degree of obstruction and establishing small-bowel normality supports the use of this technique for the further examination of patients in either group [8]. The grouping of patients with high-grade partial or complete small-bowel obstruction into a category of higher grade obstruction is based on the premise that no substantive differences exist in the imaging evaluation or clinical management of these two patient groups [2, 7]. CT is sensitive and highly accurate for the diagnosis of severe grades of small-bowel obstruction and is a technically optimal method of examining these patients [1, 37]. If abdominal radiography is to function as a decision tool or triage point for the direc-

AJR:176, January 2001

171

Lappas et al.
Statistical Association of Abdominal Radiography Findings for Distinguishing Higher Grade from Lower Grade Small-Bowel Obstruction Finding Strongly associated ( p < 0.05) Maximum small-bowel diameter Mean airfluid level width Presence of differential airfluid levels No. of gas-filled small-bowel loops > 2.5 cm Mean small-bowel diameter No. of fluid-filled small-bowel loops > 2.5 cm Stretch sign Maximum airfluid level width No. of gas- or fluid-filled small-bowel loops < 2.5 cm Maximum differential airfluid level height Mean differential airfluid level height Small bowelcolon ratio > 0.5 Marginally associated ( p 0.15) String-of-beads sign

TABLE 3

p
0.0001 a 0.0001 a 0.0003 b 0.0005 a 0.0007 a 0.0007 a 0.001 b 0.001 a 0.008 a 0.017 a 0.024 a 0.035 b 0.15 b

Note.Findings include categoric and continuous (numeric) variables. Higher grade category ( n = 41) includes high-grade partial and complete obstruction, lower grade category ( n = 40) includes low-grade partial obstruction and no obstruction. The following findings are not associated ( p > 0.3) with distinguishing grade of obstruction: distended stomach, number of airfluid levels > 10 mm, presence of colon gas or feces, presence of cecal fluid, colon segment width, maximum and mean colon diameters, small bowelto-colon ratio > 1.0, and ratio of small-bowel diameter to colon diameter.
a t test. b Fishers exact test.

TABLE 4

Threshold Values for Radiographic Findings Involving a Numeric Measurement Variable a Threshold b 2 Loops 2 Loops 25 mm 0.5 35 mm 25 mm 2 mm 2 mm

pc
0.0008 0.0003 0.003 0.035 0.00005 0.0003 0.0003 0.00005

No. of gas- or fluid-filled small-bowel loops < 2.5 cm No. of gas- or fluid-filled small-bowel loops > 2.5 cm Mean small-bowel diameter Ratio of small-bowel diameter to colon diameter Maximum airfluid level width Mean airfluid level width Maximum differential airfluid level height Mean differential airfluid level height
b Determined by tree-based modeling [12]. c Derived from Fishers exact test.

a Continuous (numeric) variables strongly associated with grade of obstruction (Table 3).

TABLE 5

Association of the Two Most Significant Abdominal Radiography Findings with Category of Small-Bowel Obstruction Neither Present One Present No. 18/41 13/40 31/81 % 44 33 38 Both Present No. 18/41 3/40 21/81 % 44 8 26

Category No. Higher grade Lower grade Total 5/41 24/40 29/81 % 12 60 36

Note.Two most significant findings are presence of differential airfluid level and mean airfluid level width 25 mm. Higher grade category (n = 41) includes high-grade partial and complete obstruction, lower grade category ( n = 40) includes lowgrade partial obstruction and no obstruction.

tion of radiologic imaging of patients suspected of small-bowel obstruction, then documenting the radiographic ndings that best distinguish between higher grade and lower grade obstruction would be of value. Therefore, our study was designed to correlate the occurrence and measurement of all potential ndings on abdominal radiographs with the severity of small-bowel obstruction as it was established in clinical cases by an accurate methodenteroclysisusing specic criteria for categorization. In an ideal situation, all enteroclysis examinations would be performed on the same day as the patients clinical presentation and admission abdominal radiography to optimize the concordance between the presenting radiographic ndings and the patency of the small-bowel lumen as determined by enteroclysis. Although most enteroclysis examinations were performed with reasonable clinical promptness, in some patients any degree of small-bowel abnormality may have resolved or perhaps even worsened by the time of enteroclysis. A prospective study might have provided an ideal analysis. Valid correlations also require an objective recording of the radiographic ndings. Abdominal radiograph interpretations were therefore conducted randomly and by a consensus of observers who were unaware of the corresponding enteroclysis diagnosis. In some cases, particularly patients with radiographs showing numerous distended smallbowel loops, a precise accounting of the number of loops and the measurement of all loop diameters could not be made with absolute certainty. Determining the margins of predominantly uid-lled small-bowel loops for diameter measurement is more difcult than determining the margins of gas-lled loops, and measurement of uid-lled loops could be underestimated in comparison. Appreciating the presence of differential air uid levels can also be difcult in cases with overlapping bowel loops and these levels are potentially misrepresented, probably underestimated, when interpreting upright abdominal radiographs. In most instances, however, we believe that scoring the various ndings was straightfoward and seldom problematic with careful scrutiny of the radiographs. The extensive data collected, the calculations of mean measurements in addition to individual measurements, and an equivalent number of radiographs being scored from the different groups, are factors that may minimize the effect that potential inaccuracies in data collection would have on results.

172

AJR:176, January 2001

Abdominal Radiography of Small-Bowel Obstruction In our clinical settinga large primary care and tertiary referral centerthe number of symptomatic patients encountered with either low-grade partial obstruction or no obstruction is considerably greater than the number of patients having a high-grade partial or complete small-bowel obstruction. Given the quantity of the data collected and analyzed, the arbitrary enrollment of an equivalent number of patients in each study group was considered a reasonable approach to time and cost considerations of the study and a satisfactory approach for valid statistical comparisons between the groups. The aim of the statistical analysis of the data was to determine from the large number of scored radiographic ndings the most signicant ndings that might be used to differentiate the grade of small-bowel obstruction. Initially, a critical p value of 0.3 was chosen to limit further analysis to the ndings (variables) that showed an association with obstruction category. A lower cut-off p value, such as the traditional 0.05, was not used because we did not want to prematurely eliminate radiographic ndings (variables) that were potentially important once other variables were taken into consideration. In essence, the value of the univariate screening process was to identify ndings likely to be signicant while reducing the number of total variables for the nal statistical analysis. Because the use of radiographic ndings characterized by a numeric measurement (termed a continuous variable in this analysis) would be difcult to implement clinically, it was necessary to transform these variables. The generation of cut-points by tree-based modeling accomplishes an appropriate modication of the continuous variables by dening a specic threshold value which, by its presence or absence, represents a more clinically usable nding to distinguish between the higher grade and lower grade obstructions. In the nal analysis, a logistic regression model was used to determine the most signicantly predictive radiographic ndings. Logistic regression is appropriate for identifying predictors when the dependent variablein this study, the degree of obstructionis in itself categoric in nature. All relevant variables with an initial p value of less than 0.3 were included in the model and then were removed individually if they did not contribute signicantly to the prediction of the obstruction. For example, it would be unlikely for the maximum and mean airuid level widths to both be necessary as predictors. If both are necessary, the logistic regression model would use both; otherwise, the model would choose the more signicantly informative one. After all noninformative variables were eliminated, the resulting version of the statistical model was used for prediction of the grade of small-bowel obstruction. Results of this study show that abdominal radiographs can provide a distinction between these groups of patients in most cases. Of the 25 radiographic ndings studied, several anticipated ndings were by association signicant in allowing distinction between higher and lower grades of obstruction such as smallbowel distention, with the maximally dilated loops averaging 36 mm in diameter and exceeding a size ratio of 0.5 with the colon, and a nearly 2.5 times increase in the number of distended loops in the abdomen. These ndings could be expected, given the nature of bowel obstruction, in that abnormal distention of either gas-lled or uidlled intestine, or both, occurs proximally to the site of obstruction and with relative collapse of the more distal intestine. In particular, two ndings derived from the upright abdominal radiograph were found to be the most signicant and most predictive of the higher grades of small-bowel obstruction: the presence of differential airuid levels and a mean width of airuid levels measuring greater than or equal to 25 mm. Differential airuid levels were present in 56% of all patients and were observed with greater frequency on the radiographs of patients in the higher grade than in the lower grade category of obstruction, 76% versus 35%, respectively. The measured differential height of these air uid levels was also notably increased with the greater severity of obstruction. We did not show in these data a signicant threshold height of the differential airuid levels as a discriminating diagnostic nding, as has been reported by others [15]. In 1993, Harlow et al. [15] showed that the specicity for the diagnosis of mechanical bowel obstruction increased, and diagnostic sensitivity decreased, as the threshold height dening a differential air uid level increased. A critical differential air uid level height of greater than or equal to 20 mm was associated with a high positive predictive value (88%) for the diagnosis. Although the study by Harlow et al. involved a radiographic comparative analysis of 62 episodes of mechanical bowel obstruction with 38 episodes of adynamic obstruction (ileus) as determined largely by clinical information, our study involved only two patients with a clinical diagnosis of an adynamic obstruction. Perhaps these differences in the patient populations being compared, including differences in clinical presentation, account for the differences observed regarding a signicant predictive height of differential airuid levels. That a critical width of airuid levels was found to be highly signicant likely reects its association with the greater degree of abnormal distention of the small-bowel loops seen in higher grades of obstruction. Long airuid levels observed in cases of obstruction have also been thought to correlate with the amount of luminal uid present, which is especially increased in obstructed loops of distal jejunum or ileum [16]. In our study, we did not determine the anatomic level of the small-bowel obstruction, so we are unable to additionally comment on this feature. Corroborating the presence of distended loops with copious luminal uid was the signicant occurrence of the stretch sign evident on supine radiographs. Interestingly, airuid levels of small width (<10 mm) or the radiographic string of beads, was only marginally associated with obstruction severity. Although this sign is considered an important indicator of the presence of uidlled bowel, its frequent presence in both higher and lower grades of obstruction suggests that its occurrence is perhaps independent of luminal distention or is quite readily perceived on upright radiographs. Considerable material has been published about the sensitivity of various radiographic ndings in the diagnosis of small-bowel obstruction, particularly in regard to airuid levels and the importance of the upright abdominal radiograph. The early observations of Frimann-Dahl [17] on the pertinence of air uid levels in distinguishing adynamic obstruction or ileus (nonobstructive distension) and mechanical obstruction lacked consistent substantiation in subsequent literature. Some studies showed the relative nonspecicity and insensitivity of these ndings in establishing the diagnosis of small-bowel obstruction, although the signicance of airuid levels as an indicator of small-bowel luminal stasis is widely accepted [1821]. The presence of differential airuid levels on abdominal radiographs has also been regarded as an insensitive diagnostic nding in differentiating the mechanical versus functional nature of an intestinal obstruction because of their potential occurrence in both of these conditions [15]. Other researchers have even recommended that the upright radiograph be eliminated from

AJR:176, January 2001

173

Lappas et al. the acute abdominal radiographic series because of its inability to substantially alter the sensitivity of the examination in the diagnosis of small-bowel obstruction [22, 23]. However, in those studies the patient populations were nonuniform and included patients with a wide variety of abdominal complaints and conditions such as abdominal pain, adynamic ileus, acute surgical abdomen, small-bowel obstruction, and colon obstruction. Few patients with a diagnosis of small-bowel obstruction were involved in the one study [22], whereas in the other the upright and supine radiographs were not evaluated independently of each other [23]. Citing such limitations and further recognizing the need to render as rapid and accurate a diagnosis as possible, other researchers have maintained support for the value of upright abdominal radiographs [24]. Unlike these previous reports, our analysis focuses on an issue other than diagnostic sensitivity: the relevance of radiographic ndings in distinguishing the severity of small-bowel obstruction, because we think this distinction has important ramications for the practical imaging examination of symptomatic patients in whom obstruction is clinically suspected by the ndings of physical examination and pertinent history. The two radiographic ndings considered of greatest signicance as predictors of the severity of obstruction were ndings seen only on the upright abdominal radiograph. When both of these radiographic ndings are considered, their combined presence or absence in any given case is a strong positive (86%) or negative (83%) indicator of the degree of patency of the small-bowel lumen. Although upright radiographs alone may not be particularly sensitive in the detection of small-bowel obstruction, they may be of value in distinguishing patients with either a highgrade or a complete obstruction from those with a low-grade partial obstruction or those who are symptomatic but have no obstruction. The information obtained from the upright radiograph makes it an integral part of abdominal radiographic examinations in patients with suspected small-bowel obstruction by helping to direct, if needed, the additional imaging method that is appropriate for further workup. The results of this study validate prior recommendations to use abdominal radiographs as a practical tool in the radiologic management of small-bowel obstruction [2]. This issue is important because the erroneous application of diagnostic procedures not only adds to health care costs but also may delay diagnosis and adversely affect patient treatment decisions.
Acknowledgment
10. Walsh DW, Bender GN, Timmons JH. Comparison of computed tomography-enteroclysis and traditional computed tomography in the setting of suspected partial small bowel obstruction. Emerg Radiol 1998;5:2937 11. Markus JB, Somers S, Slobodan EF, Moola C, Stevenson GW. Interobserver variation in the interpretation of abdominal radiographs. Radiology 1989;171:6971 12. Clark LA, Pregibon D. Tree-based models. In: Chambers JM, Hastie TJ, eds. Statistical models in S. Pacic Grove, CA: Wadsworth & Brooks/Cole Advanced Books and Software, 1992:377390 13. Hosmer DW, Lemeshow S. Assessing the t of the model. In: Hosmer DW, Lemeshow S, eds. Applied logistic regression. New York: Wiley, 1989:135173 14. Hosmer DW, Lemeshow S. Special topics: polytomous logistic regression. In: Hosmer DW, Lemeshow S, eds. Applied logistic regression. New York: Wiley, 1989:216245 15. Harlow CL, Stears RLG, Zeligman BE, Archer PG. Diagnosis of bowel obstruction on plain abdominal radiographs: signicance of air-uid levels at different heights in the same loop of bowel. AJR 1993;161:291295 16. Bryk D, Wolf BS. A radiological evaluation of small bowel activity in the acute abdomen. Crit Rev Diagn Imaging 1977;10:99128 17. Frimann-Dahl J. Special pathological ndings: stenosis and obstruction. In: Frimann-Dahl J, ed. Roentgen examinations in acute abdominal diseases, 3rd ed. Springeld, IL: Thomas, 1974:88136 18. Gammill SL, Nice CM Jr. Air-uid levels: their occurrence in normal patients and their role in the analysis of ileus. Surgery 1972;71:771780 19. Levin B. Mechanical small bowel obstruction. Semin Roentgenol 1973;8:281297 20. Schwartz SS. The differential diagnosis of intestinal obstruction. Semin Roentgenol 1973;8:323338 21. Baker SR, Cho KC. Plain lm radiology of the intestines and appendix. In: Baker SR, Cho KC, eds. The abdominal plain lm with correlative imaging, 2nd ed. Stamford, CT: Appleton & Lange, 1999:217367 22. Mirvis SE, Young JWR, Keramati B, McCrea ES, Tarr R. Plain lm evaluation of patients with abdominal pain: are three radiographs necessary? AJR 1986;147:501 503 23. Simpson A, Sandeman D, Nixon SJ, Goulbourne IA, Grieve DC, Macintyre IM. The value of the erect abdominal radiograph in the diagnosis of intestinal obstruction. Clin Radiol 1985;36:4142 24. Mindelzun RE, McCort JJ. Denition of acute abdominal series. (answer to question) AJR 1996; 166:716718

We thank Mark P. Hanna of the Division of Biostatistics, Department of Medicine, Indiana University School of Medicine for the statistical support and analysis involved in this study.

References
1. Maglinte DDT, Reyes BL, Harmon BH, et al. Reliability and role of plain lm radiography and CT in the diagnosis of small-bowel obstruction. AJR 1996;167:14511455 2. Maglinte DDT, Balthazar EJ, Kelvin FM, Megibow AJ. The role of radiology in the diagnosis of smallbowel obstruction. AJR 1997;168:11711180 3. Megibow AJ, Balthazar EJ, Cho KC, et al. Bowel obstruction: evaluation with CT. Radiology 1991;180:313318 4. Fukuya T, Hawes DR, Lu CC, Chang PJ, Barloon TJ. CT diagnosis of small bowel obstruction: efcacy in 60 patients. AJR 1992;158:765769 5. Gazelle GS, Goldberg MA, Wittenberg J, Halpern EF, Pinkney L, Mueller PR. Efcacy of CT in distinguishing small-bowel obstruction from other causes of small-bowel dilatation. AJR 1994;162:4347 6. Frager D, Medwid SW, Baer JW, Mollinelli B, Friedman M. CT of small-bowel obstruction: value in establishing the diagnosis and determining the degree and cause. AJR 1994;162:3741 7. Taourel PG, Fabre J-M, Pradel JA, Senterre EJ, Megibow AJ, Bruel J-M. Value of CT in the diagnosis and management of patients with suspected acute small-bowel obstruction. AJR 1995;165:11871192 8. Shrake PD, Rex DK, Lappas JC, Maglinte DDT. Radiographic evaluation of suspected small bowel obstruction. Am J Gastroenterol 1991;86:175178 9. Maglinte DDT, Gage S, Harmon B, et al. Obstruction of the small intestine: accuracy and role of CT in diagnosis. Radiology 1993;186:6164

174

AJR:176, January 2001