Abdominal CT as a predictor of outcome before attempted direct percutaneous endoscopic jejunostomy

ORIGINAL ARTICLE

Abdominal CT as a predictor of outcome before attempted direct percutaneous endoscopic jejunostomy John T. Maple, DO, Bret T. Petersen, MD, Todd H. Baron, MD, Gavin C. Harewood, MD, C. Daniel Johnson, MD, Grant D. Schmit, MD Rochester, Minnesota, USA

Background: Direct percutaneous endoscopic jejunostomy (DPEJ) placement succeeds in 72% to 86% of attempts. Failure is most often because of inadequate transillumination or gastroduodenal obstruction. Even in failed cases, patients are exposed to the risks of anesthesia, exploratory percutaneous needle punctures, and the cost burden of suboptimal resource utilization. Hence, a preprocedure predictor of outcome would be useful. Objective: To evaluate whether review of clinically available abdominal CTs can predict the outcome of subsequent DPEJ attempts. Design: Retrospectively conducted blinded review of abdominal CTs performed within 30 days before attempted DPEJ. Objective anatomic features potentially pertinent to DPEJ success were scored, and a prediction of the anticipated procedural outcome was made. Setting: A large tertiary referral center. Patients: A total of 115 patients who underwent attempted DPEJ and who also had an abdominal CT in the preceding 30 days. Main Outcome Measurements: Reviewer’s overall prediction of success, 3 objective anatomic measurements. Results: For the overall prediction of success, a CT performed poorly, with a sensitivity of 60%, a specificity of 53%, a positive predictive value of 71%, and a negative predictive value of 40%. Mean abdominal-wall thickness was significantly greater in the failures than the successes (27 vs 21 mm, P Z.02), and just 39% of the procedures in patients with an abdominal-wall thickness O3 cm were successful. Limitations: Retrospective. Conclusions: Failed DPEJ attempts were associated with greater patient abdominal-wall thickness, and this should be taken into consideration before attempted DPEJ. Otherwise, review of existing abdominal CTs appears to have limited utility in predicting DPEJ outcome. (Gastrointest Endosc 2006;63:424-30.)

Direct percutaneous endoscopic jejunostomy (DPEJ) was first described by Shike et al1 in 1987. DPEJ provides more reliable enteral delivery than a gastrostomy tube with jejunal extension,2 and the number of referrals for DPEJ is increasing at our institution. Success rates for DPEJ placement have ranged from 72% to 100%,1-7 with the largest published series (n Z 150) reporting successful placement in 86%.3 The reasons for unsuccessful attempts at DPEJ cited most often are the inability to identify an adequate insertion site via transillumination Copyright ª 2006 by the American Society for Gastrointestinal Endoscopy 0016-5107/$32.00 doi:10.1016/j.gie.2005.10.035

through the jejunal and abdominal wall (w70% of failed DPEJs) and the inability to intubate the jejunum because of gastric outlet or proximal small-bowel obstruction (w30%).1-7 Small-bowel endoscopy (or push enteroscopy) with an unsuccessful DPEJ placement is not an entirely benign procedure. Patients are exposed to the inherent risks of upper endoscopy, which relate both to sedation and analgesia, as well as to the procedure itself. The bulk of adverse events from conscious sedation are cardiopulmonary in nature, such as myocardial ischemia, arrhythmias, hypoxemia, and aspiration.8,9 A retrospective review by the American Society for Gastrointestinal Endoscopy and the Food and Drug Administration of over 21,000

424 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 3 : 2006

www.giejournal.org

Maple et al

endoscopic procedures reported a cardiopulmonary complication rate of 5.4 per 1000 and a death rate of 0.3 per 1000.10 However, the risk for arterial desaturation (and thus cardiopulmonary events) from sedation increases with advancing age and the number of comorbid illnesses.11 The population of patients undergoing DPEJ is an older, sicker population than those undergoing routine EGD, and, thus, risks are greater in this population. Perhaps reflecting this, a series of 314 consecutive PEG-tube placements was associated with a 1% mortality rate, with all 3 deaths associated with aspiration or laryngospasm.12 Furthermore, even limited enteroscopy to seek out a site for DPEJ takes more time than a routine EGD or PEG, subjecting patients to a more prolonged period of sedation. Procedure-related complications, such as bleeding, perforation, and infection, occur rarely at EGD but are more common when exploratory percutaneous needle punctures are performed during attempted DPEJ and certainly can have severe consequences.8 Data from our own institution13 noted 1 death related to profound bleeding from a site in the jejunal mesentery after an unsuccessful trocar pass at attempted DPEJ. Beyond patient safety, unsuccessful procedures engender both a cost burden and suboptimal resource utilization. A method for predicting success or failure of DPEJ before the procedure would be desirable. While use of US to locate a jejunal loop during a DPEJ placement has been described,14 the use of imaging before DPEJ to help determine feasibility has not. Abdominal CT is a widely available tool that can afford a clear picture of the anatomic position of the proximal jejunal loops in relation to the abdominal wall and other structures and also, in many cases, can adequately demonstrate gastric or duodenal obstruction (Fig. 1). Many patients scheduled for DPEJ have significant comorbidities and a history of multiple imaging studies in prior months. We sought to evaluate whether a review of recent clinically indicated abdominal CTs would be predictive of subsequent success or failure of DPEJ placement.

CT prediction of DPEJ placement success

Capsule Summary What is already known on this topic d

DPEJ placement succeeds in 72% to 86% of attempts, and failure is most often caused by inadequate transillumination or gastroduodenal obstruction.

What this study adds to our knowledge d

The overall prediction of DPEJ success based on a review of preprocedure abdominal CTs was not strongly correlated with the actual outcome, although fewer than 40% of DPEJ placement attempts succeeded in patients with an AWT greater than 3 cm.

Two large institutional databases were searched to identify 316 attempted DPEJ placements at Mayo Clinic Rochester between January 1996 and August 2004. All but 9 patients consented to have their medical records used for research. The primary investigator then reviewed the electronic medical records of these patients to ascertain whether or not the patients had undergone abdominal CT in the 30 days preceding the attempted DPEJ placement; 124 such cases were identified. After eliminating multiple procedures on the same patient, DPEJs that failed for reasons other than lack of transillumination or gastroduodenal obstruction, and DPEJs enabled only by dilating and/or

stenting a duodenal obstruction, 120 patients remained: 79 successes and 41 failures. Three cases were selected as ‘‘teaching cases,’’ and 2 other cases were randomly eliminated to arrive at the study group of 115 cases. Four gastroenterologists (2 who perform DPEJ and 2 who do not) and 2 radiologists with a focus on GI and abdominal imaging participated in the study. Initially, all participants reviewed 4 abdominal CTs (2 successes, 2 failures) in an unblinded fashion on attempted DPEJ patients as part of a prepared tutorial, with associated written commentary provided by the senior investigator (B.T.P.), highlighting features of each scan that seemed to portend the observed outcome. The reviewers were given the following points to consider when reading the CTs: d Is the duodenum obstructed by a mass or a stricture? d Does the first or second loop of jejunum lie close to the abdominal wall or deep? d If not, is there a more distal jejunal loop that lies favorably close to the abdominal wall? d Where the accessible jejunum lies closest to the abdominal wall, are there any overlying structures (eg, ribs, colon, other small-bowel loop) that may impair access? d Is the patient significantly obese? d Is there significant ascites? Each reviewer then viewed and interpreted 25 scans (7 shared cases and 18 unique cases) on a computer workstation reserved for research purposes and not linked to the electronic medical record. Reviewers identified case films by clinic number and thus were blinded to the outcome of the subsequent DPEJ. The data entry sheets prompted each reviewer to predict the likelihood of success or failure of DPEJ placement on a 1 to 4 scale, corresponding to the descriptors ‘‘very likely, probably, probably not, and very unlikely.’’ If failure was predicted, reviewers assigned a reason for probable failure: lack of transillumination, gastric/duodenal obstruction, both, or other. A software-based measuring tool was used to measure both the abdominal-wall thickness (AWT) and the total skin-to-jejunum distance at the level of the most promising site. Reviewers also estimated the jejunal

www.giejournal.org

Volume 63, No. 3 : 2006 GASTROINTESTINAL ENDOSCOPY 425

PATIENTS AND METHODS Study design

CT prediction of DPEJ placement success

Maple et al

Figure 1. Favorable vs unfavorable anatomy for DPEJ. A, Favorable DPEJ anatomy. The yellow line demonstrates the path of the duodenal sweep and the proximal jejunum. After a soft turn just past the ligament of Treitz, the jejunum makes a straight path to the anterior abdominal wall in the left upper quadrant. This DPEJ placement was successful and uneventful. B, Unfavorable DPEJ anatomy. The yellow circle illustrates a solid/cystic mass in the 3rd and 4th parts of the duodenum in this patient with lymphoma. Clinically, the endoscopist had great difficulty bypassing this mass, and was only able to traverse a short distance beyond it. Adequate transillumination/finger indentation was not achieved. C, Unfavorable DPEJ anatomy. In this patient with gastroparesis, the 2nd portion of the duodenum is shown with an arrowhead. The proximal jejunum traveled back across the midline just after the ligament of Treitz (arrows), unfortunately, deep to the transverse colon (encircled) and poorly opposed to the abdominal wall, as shown. Clinically, adequate transillumination/finger indentation was never achieved, despite prolonged efforts.

loop number beyond the ligament of Treitz for optimal insertion and predicted the anatomic location on the abdominal wall for DPEJ. Finally, reviewers rated the quality and the adequacy of the clinical pre-DPEJ CTs for making these assessments on a 1 to 4 scale.

Rochester, performed the procedures on the patients in this study. Three endoscopists accounted for 92 of the placement attempts, and the remaining 21 placement attempts were performed by 9 other endoscopists.

Sample-size assessment Procedure methods We use a modified Ponsky pull-through method for DPEJ placement by using a standard 20F pull-method PEG kit, as detailed in full by Rumalla and Baron.5 Twelve gastroenterologists, with extensive experience with percutaneous endoscopic gastrostomy and who routinely staff a ‘‘complex endoscopy’’ schedule at Mayo Clinic

Predicting a successful outcome with DPEJ every time will be correct 70% of the time, given that 70% of DPEJ placements succeed. Given that a 90% rate of correct prediction would be clinically useful, at least 42 patients will be required to detect a 20% difference between the predicted outcome after abdominal CT review and the actual DPEJ outcome, with 80% power at a significance level of

426 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 3 : 2006

www.giejournal.org

Maple et al

CT prediction of DPEJ placement success

TABLE 1. Patient demographics Successes

TABLE 2. Mechanisms for DPEJ placement failure* Failures

No.

75

38

Mean age

59

64.7

Total 113 60.9

No. (%) Lack of adequate transillumination/finger indentation Absent

Gender

Suboptimal, with an unsuccessful needle or trocar pass

31 (82%) 24 (63%) 7 (18%)

Men

46

17

63

Women

29

21

50

Unable to pass scope to jejunum

5 (13%)

2

7

Difficulty passing scope C lack of transillumination

2 (5%)

5

Regional*

41

21

62

Total

Distant (all others)

29

15

44

*Adapted from Maple JT, et al. Direct percutaneous endoscopic jejunostomy: outcomes in 307 consecutive attempts. Am J Gastroenterol 2005;100:2681-8.

Gastroparesis

11

8

19

Cancerz

34

13

47

Prior surgeryx

10

2

12

1

5

6

19

10

29

Residence Olmsted County

Indication for DPEJy

High aspiration risk Other

*Minnesota other than Olmsted County, Wisconsin, and Iowa. yIndications for jejunal feeding rather than gastric feeding. zMost commonly esophageal or gastric. xSurgically altered anatomy, precluding PEG placement.

38 (100%)

pair of reviewers for the overall prediction of outcome in the 7 cases rated by all 6 observers; the reported kappa statistic is an average of all pairwise measurements.

RESULTS Patient demographics and procedure characteristics

JMP statistical software (SAS Institute Inc, SAS Campus Drive, Cary, NC) was used for the analysis. A P value !.05 was considered statistically significant for all tests. The responses for overall prediction of success were dichotomized into a binary (yes/no) prediction and were compared with actual outcome; the resulting 2
2 table allowed calculation of sensitivity, specificity, and positive and negative predictive values. Mean AWT, mean skin-tojejunum distance, and mean loop depth (calculated from the former 2 measurements) were analyzed as continuous variables against the actual outcome by using t tests and Wilcoxon rank sum tests when appropriate. Categorical variables with 2 groups were compared with the Fisher exact test, while categorical variables with more than 2 groups were compared by using a c2 test. The optimal cutoff value for AWT as a predictor of success was selected after the generation of a receiver-operating characteristics curve for these variables. For the 7 cases reviewed by all 6 reviewers, median responses for categorical and ordinal variables were used, and mean values for continuous variables were used when analyzing data for the study group. Interobserver correlation was measured with an unweighted Cohen’s kappa for the agreement between each possible

Pertinent demographics for the study group, including DPEJ placement success and indications, are shown in Table 1. Of the 115 study subjects, CTs were reviewed in 113. One patient’s electronic film file was corrupted and was unable to be viewed; another was dropped from review when the interpreting clinician recognized the clinical case. In this highly selected study group, 75 of 113 (66%) of the DPEJs were successfully placed. However, in comparison to the larger group of 307 attempts from which these cases were selected, the placement success of those patients with a CT in the prior 30 days was not significantly different from those who had not undergone a CT in the prior 30 days.13 Mechanisms for failure in the 38 unsuccessful DPEJ attempts are shown in Table 2. Adverse events (AE) associated with DPEJ placement and removal in these patients were previously reported in detail as part of a larger series.13 In this study group, 26 AEs were observed in 24 of 113 (21.2%) of the cases; 2 patients experienced 2 AEs. Most AEs were minor, including 10 site infections that required oral antibiotics and 6 instances of persistent PEJ site pain beyond 72 hours. However, 4 severe AEs were observed: 1 major PEJ site bleed required multiple therapeutic endoscopies, 2 bowel perforations, and 1 jejunal volvulus. The remaining AEs were of moderate severity and included 2 enterocutaneous fistulas, 2 cases of severe site pain that required tube removal or hospitalization for intravenous analgesia, 1 jejunal pressure ulcer/partial buried bumper, and 1 cutaneous site abscess.

www.giejournal.org

Volume 63, No. 3 : 2006 GASTROINTESTINAL ENDOSCOPY 427

5% (a Z 0.05) when using a 2-sided 1-sample test for proportions.

Statistical methods

CT prediction of DPEJ placement success

Maple et al

TABLE 3. Performance characteristics of abdominal CT for the overall prediction of success or failure of subsequent DPEJ Actual DPEJ outcome* Success

Failure

TABLE 4. Objective measurements of distance correlated with DPEJ outcome Failures

Successes

P value

Mean AWT, mm

26.7

20.9

.018

Mean skin-to-jejunum distance, mm

40.5

33.2

.037

Mean jejunal loop depth from the abdominal wall, mm

13.7

12.2

NS

Total

CT-predicted DPEJ outcome Success

45

18

63

Failure

30

20

50

Total

75

38

113

*Sensitivity, 60%; specificity, 53%; positive predictive value, 71%; negative predictive value, 40%.

Subjective prediction of success The overall prediction of procedural success by the clinical CT reviewers was not strongly correlated with the actual DPEJ outcome, as shown in Table 3. Used in this manner, CT-guided prediction had a sensitivity of 60% and a specificity of 53%. The performance of the overall prediction of outcome remained poor when reevaluated after controlling for reader specialty, quality of CT, and predicted reason for DPEJ failure. Additionally, the accuracy of overall prediction of outcome by reviewers among the 38 failed DPEJ attempts was not different between those cases that failed because of inadequate transillumination and those that failed because of gastroduodenal obstruction. Lastly, no single reader was significantly more accurate in his predictions than the other reviewers.

Objective distance measurements

the patients with an AWT O3 cm, 12 of 14 (86%) failed because of inadequate transillumination compared with 19 of 24 (79%) for those with an AWT %3 cm (P Z not significant [NS]).

Other predictors Neither the estimated number of jejunal loops from the ligament of Treitz to the optimal placement site, nor the predicted anatomic location for placement on the abdominal wall was significantly different between the group of successful DPEJ placements and failures.

Interobserver correlation Correlation between the 6 reviewers was measured in regard to the overall subjective prediction of success or failure in the 7 shared cases. Reviewer agreement was not strong, with a kappa of 0.11.

DISCUSSION

The correlation between 3 objective measurements of distance on the abdominal CTs with the outcome of the procedure is shown in Table 4. Mean AWT measured on CT was significantly greater in patients whose DPEJ attempt failed than in those who had a successful DPEJ placement (26.7 vs 20.9 mm, P Z.018). A receiver operating characteristics curve (not shown) was generated for AWT vs procedure outcome. This demonstrated that the best performance cutoff occurred at 30 mm (3 cm). Using this value produced a sensitivity of 88% (just 12% of successful DPEJ placements were in patients with an AWT O3 cm), though an AWT %3 cm had poor specificity for outcome. These data are summarized in Table 5. The higher rate of placement success in those patients with an AWT %3 cm (73%) vs those with an AWT O3 cm (39%) was strongly statistically significant (P Z .003). PEJ placement failed in all 7 patients with an AWT O4 cm (P ! .001). An abdominal CT slice from a patient with an AWT O4 cm in whom DPEJ placement was unsuccessful is demonstrated in Figure 2. As Table 2 demonstrates, a significant majority (31/38 [82%]) of the DPEJ placement failures for the entire study group were caused by inadequate transillumination. Of

In this study, we sought to determine the predictive value of clinically available abdominal CTs in forecasting the outcome of subsequent DPEJ based on favorable or unfavorable anatomy. For the overall prediction of success by clinical reviewers, CT did not perform well. However, we observed that less than 40% of DPEJ placement attempts in patients with an AWT O3 cm succeeded. Thus, while failed DPEJ attempts occur in patients with a broad spectrum of AWTs and an AWT %3 cm certainly does not guarantee success, our data suggest that having an AWT O3 cm is a marker for a lower likelihood of DPEJ placement success. Though patients in need of jejunal feeding are not often overweight, a subset of such patients exists. It stands to reason that a larger amount of subcutaneous adipose tissue hinders adequate transillumination, finger indentation, and transcutaneous jejunal puncture, and thus clinicians may be less inclined to attempt a percutaneous needle puncture for fear of intervening structures (eg, colon). Thus, the findings of this study that the mean AWT was significantly greater in failed DPEJ patients than successes and that none of the 7 DPEJs in patients with an AWT R4 cm was successful, have a rational clinical basis. These findings should be considered as patients are

428 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 3 : 2006

www.giejournal.org

Maple et al

CT prediction of DPEJ placement success

TABLE 5. DPEJ outcome as a function of AWT DPEJ outcome* Success

Failure

Total

%3 cm

66

24

90

O3 cm

9

14

23

75

38

113

AWT

Total

*Sensitivity, 88%; specificity, 37%; positive predictive value, 73%; negative predictive value, 61%.

being clinically evaluated for possible DPEJ if an abdominal CT is available. While subcutaneous adipose tissue on abdominal CT has been shown to correlate with other noninvasive anthropometric measures, eg, waist circumference,15 the direct relationship of waist circumference (or other measures of obesity, such as body mass index or waist-hip ratio) with DPEJ outcome has not been demonstrated. Such measurements were unfortunately not available in the majority of our study patients. The ‘‘loop depth’’ is simply the calculated difference between the AWT and the total skin-to-jejunal loop distance. Because mean loop depth alone had no association with the observed outcome, it appears that the significance of the mean total skin-to-jejunal loop distance is solely a function of the AWT. Anecdotal experience suggests that the use of extra-long needles, such as those designed for lumbar puncture, may facilitate jejunal puncture in such cases. However, given concerns that this technique probably increases the likelihood of inadvertently puncturing an intervening structure (because transillumination and indentation are often suboptimal in these cases), we have not adopted this technique into our practice. Thus, given the poor performance of endoscopic PEJ placement in this population of patients with a significant amount of subcutaneous abdominal fat, referral for surgical jejunostomy (SJ) is a reasonable alternative to attempted DPEJ. The complication profiles of DPEJ and SJ are difficult to compare because of heterogenous patient populations and the paucity of complications data about patients operated on solely for jejunal access (more commonly SJ is performed with another operation). However, adverse events appear to be similar with these techniques: rare mortality (typically !1%) has been reported with both procedures, and similar rates of serious events, such as bowel perforation (%2.5%) and jejunal volvulus (%1%), have been reported.3,13,16-18 This similarity supports the conclusion that, in patients at high risk for placement failure at DPEJ, surgical jejunostomy should be considered if the clinical situation warrants. We evaluated whether procedure time could confound the observed association between AWT and success (e.g., endoscopists gave up more quickly in heavier patients).

Among all 113 cases, there was no correlation between AWT and procedure time (Pearson R-square coefficient, 0.03). Notably, in 39 of 113 cases, a procedure other than attempted DPEJ was also performed (eg, esophageal stent or PEG-J in a failed DPEJ attempt), and this occurred more commonly in failed DPEJ attempts. However, confining this calculation to those 74 cases in which attempted DPEJ was the only procedure performed did not change the lack of correlation between procedure time and AWT (Pearson R-square coefficient, 0.08). Thus, procedure time did not appear to be a confounder of the effect of AWT on DPEJ placement success. A number of considerations may explain why the primary outcome (overall CT-based prediction of success) was not accurate. Clinical experience has shown that occasionally good transillumination for DPEJ is seen on withdrawal that was not apparent on insertion or that a failed procedure because of inadequate transillumination may succeed the following day. However, in our clinical experience, there also are some patients who have undergone attempted DPEJ on 3 separate occasions, and good transillumination was never achieved despite prolonged efforts. Thus, the extent to which endoscopic manipulation of the bowel can overcome seemingly unfavorable jejunal anatomy for DPEJ (such as overlying colon or ribs, or very deep-lying loops) by ‘‘driving’’ the bowel to the abdominal wall is not well established, but it may be a factor. Additionally, it is unclear how durable and thus how meaningful CT images of small-bowel anatomy are, in that bowel loops may have changed location in the days to weeks between the imaging and the attempted DPEJ in this study. The most common error by reviewers in this study was predicting failure in procedures that eventually succeeded. Speculation can be made that the endoscope was frequently able to alter the position of the jejunum

www.giejournal.org

Volume 63, No. 3 : 2006 GASTROINTESTINAL ENDOSCOPY 429

Figure 2. Increased AWT. This 54-year-old female patient had severe nausea and vomiting, with delayed scintigraphic gastric emptying and antral hypomotility on gastroduodenal manometry. Adequate transillumination was not achieved at attempted DPEJ placement. This CT image demonstrates increased AWT, which undoubtedly hindered the procedure.

CT prediction of DPEJ placement success

and maneuver away from overlying structures or to reduce an unfavorably deep position in the abdominal cavity. Additionally, in 19 of 30 successful cases that were predicted to fail, the optimal insertion site was estimated to be 3 or more jejunal loops beyond the ligament of Treitz. This indicates that the reviewers thought that the optimal jejunal site was fairly distal and possibly out of the reach of the endoscope in these cases. However, push enteroscopy with a pediatric colonoscope typically reaches 50 cm beyond the ligament of Treitz, probably well into a third or fourth loop of jejunum, and the telescoping of jejunum onto the colonoscope may even underestimate the distance traversed. Additionally, while the pediatric colonoscope is routinely used as the initial instrument for DPEJ, endoscopists in our institution occasionally change to an enteroscope, with or without an overtube, if transillumination fails with the colonoscope. Thus ‘‘distal’’ sites may be more accessible than the clinical CT reviewers appreciated. In regard to reviewer agreement: for the 7 scans interpreted by all 6 readers, the overall prediction of outcome was unanimously interpreted in 2 cases, had agreement among 5 of 6 readers in 2 cases, and agreement in 4 of 6 readers in 2 cases. These outcomes are summarized by the kappa coefficient for inter-rater agreement, which was 0.11 in this case, a value not associated with good concordance. However, it should be noted that no single reader was significantly more accurate in his predictions than the other reviewers. Thus, though concordance was not tightly shown per individual scan among the reviewers, the performance of each reviewer in predicting a group of 25 scans was similar and, in all cases, not accurate enough to make CT-based predictions a clinical recommendation. This study is limited by its retrospective nature and is thus hypothesis generating, and the findings should be replicated before firm conclusions are drawn. It should be noted that this study used clinically available abdominal CTs ordered for other purposes. It is possible that protocolized CTs based on this indication, eg, CT enterography, may yield better results, though certainly at some expense. Also, the study population that had undergone CT scanning had an overrepresentation of cancer diagnoses compared with the general PEJ referral population. This may hinder the generalizability of our findings. In summary, we have demonstrated that abdominal CTs, when clinically available, may have some utility in patient selection for DPEJ, in that the procedure was more likely to fail in patients with greater AWT, particularly when this measure was more than 3 cm. Beyond this, however, an abdominal CT appears to have limited utility in predicting DPEJ outcome.

Maple et al

REFERENCES 1. Shike M, Schroy P, Ritchie MA, et al. Percutaneous endoscopic jejunostomy in cancer patients with previous gastric resection. Gastrointest Endosc 1987;33:372-4. 2. Fan AC, Baron TH, Rumalla A, Harewood GC. Comparison of direct percutaneous endoscopic jejunostomy and PEG with jejunal extension. Gastrointest Endosc 2002;56:890-4. 3. Shike M, Latkany L, Gerdes H, Bloch AS. Direct percutaneous endoscopic jejunostomies for enteral feeding. Gastrointest Endosc 1996; 44:536-40. 4. Mellert J, Naruhn MB, Grund KE, Becker HD. Direct endoscopic percutaneous jejunostomy (EPJ). Surg Endosc 1994;8:867-70. 5. Rumalla A, Baron TH. Results of direct percutaneous endoscopic jejunostomy, an alternative method for providing jejunal feeding. Mayo Clin Proc 2000;75:807-10. 6. Bueno JT, Schattner MA, Barrera R, Gerdes H, Bains M, Shike M. Endoscopic placement of direct percutaneous tubes in patients with complications after esophagectomy. Gastrointest Endosc 2003;57:536-40. 7. Barrera R, Schattner M, Nygaard S, et al. Outcome of direct percutaneous endoscopic jejunostomy tube placement for nutritional support in critically ill, mechanically ventilated patients. J Crit Care 2001;16: 178-81. 8. Chan MF. Complications of upper gastrointestinal endoscopy. Gastrointest Endosc Clin N Am 1996;6:287-303. 9. Benjamin SB. Complications of conscious sedation. Gastrointest Endosc Clin N Am 1996;6:277-86. 10. Arrowsmith J, Gerstman B, Fleischer D, et al. Results from the American Society for Gastrointestinal Endoscopy/U.S. Food and Drug Administration collaborative study on complication rates and drug use during gastrointestinal endoscopy. Gastrointest Endosc 1991;37:421-7. 11. Iber FL, Sutberry M, Gupta R, Kruss D. Evaluation of complications during and after conscious sedation for endoscopy using pulse oximetry. Gastrointest Endosc 1993;39:620-5. 12. Larson DE, Burton DD, Schroeder KW, DiMagno EP. Percutaneous endoscopic gastrostomy: indications, success, complications, and mortality in 314 consecutive patients. Gastroenterology 1987;93:48-52. 13. Maple JT, Petersen BT, Baron TH, et al. Direct percutaneous endoscopic jejunostomy (DPEJ): outcomes in 307 consecutive attempts. Am J Gastroenterol 2005;100:2681-8. 14. Sharma VK, Close T, Bynoe R, Vasudeva R. Ultrasound-assisted direct percutaneous endoscopic jejunostomy (DPEJ) tube placement. Surg Endosc 2000;14:203-4. 15. Busetto L, Baggio MB, Zurlo F, Cararro R, Digito M, Enzi G. Assessment of abdominal fat distribution in obese patients: anthropometry versus computerized tomography. Int J Obes 1992;16:731-6. 16. Gerndt SJ, Orringer MB. Tube jejunostomy as an adjunct to esophagectomy. Surgery 1994;115:164-9. 17. Holmes JH, Brundage SI, Yuen PC, Hall RA, Maier RV, Jurkovich GJ. Complications of surgical feeding jejunostomy in trauma patients. J Trauma 1999;47:1009-12. 18. Myers JG, Page CP, Stewart RM, Schwesinger WH, Sirinek KR, Aust JB. Complications of needle catheter jejunostomy in 2,022 consecutive applications. Am J Surg 1995;170:547-51.

Received April 15, 2005. Accepted October 14, 2005.

ACKNOWLEDGMENTS

Current affiliations: Department of Internal Medicine, Division of Gastroenterology and Hepatology (Drs Maple, Petersen, Baron, Harewood) and Department of Radiology (Drs Johnson, Schmit), Mayo Clinic College of Medicine, Rochester, Minnesota, USA.

We thank Felicity Boyd-Enders from the Division of Biostatistics at our institution for her review for the statistical methods.

Reprint requests: Bret T. Petersen, MD, Department of Internal Medicine, Division of Gastroenterology, Mayo Clinic College of Medicine, Charlton 8 North, 200 First St SW, Rochester, MN 55905.

430 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 3 : 2006

www.giejournal.org