j o u r n a l o f s u r g i c a l r e s e a r c h 1 8 5 ( 2 0 1 3 ) 5 8 1 e5 8 6
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.JournalofSurgicalResearch.com
Pneumatosis and portal venous gas: do CT findings reassure? Murad Bani Hani, MD,a Farin Kamangar, MD, PhD,b Sarah Goldberg, MD,c Jose Greenspon, MD,d Priti Shah, MD,e Carmine Volpe, MD,d,f Douglas J. Turner, MD,d,f Karen Horton, MD,g Elliot K. Fishman, MD,g Isaac R. Francis, MD,h Barry Daly, MD,e and Steven C. Cunningham, MDa,d,* a
Department of Surgery, Saint Agnes Hospital, Baltimore, Maryland Department of Public Health Analysis, School of Community Health and Policy, Morgan State University, Baltimore, Maryland c Department of Medicine, University of Maryland, Baltimore, Maryland d Department of Surgery, University of Maryland, Baltimore, Maryland e Department of Radiology, University of Maryland, Baltimore, Maryland f Department of Surgery, Baltimore VA Medical Center, Baltimore, Maryland g Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland h Department of Radiology, University of Michigan, Ann Arbor, Michigan b
article info
abstract
Article history:
Background: Small, single-institution studies have suggested risk factors for bowel
Received 4 February 2013
ischemia/necrosis (I/N) in patients with computed tomography (CT) findings of pneuma-
Received in revised form
tosis (PN) and portal venous gas (PVG). Here, analysis has been expanded in a large,
19 May 2013
multicenter study.
Accepted 5 June 2013
Materials & methods: Logistic regression models and receiver operating characteristic curves
Available online 29 June 2013
were used to construct a scoring system for I/N in cases of PN/PVG. Results: Of 265 patients with PN/PVG identified, 209 had adequate data. In unadjusted
Keywords:
analyses the following variables were significantly associated with I/N: age, peritoneal
Pneumatosis
signs, ascites, the presence of both PVG and PN, blood urea nitrogen (BUN), CO2, albumin,
Portal venous gas
and a history of hypertension, myocardial infarction, or stroke. In contrast, the CT findings
Intestine
of mesenteric stranding, bowel-wall thickening, and type of PN were not associated with
Scoring system
I/N. In adjusted analyses, three variables were significantly associated with I/N: age 60 y
Computed tomography
(odds ratio ¼ 2.51, 95% confidence interval: 1.26e4.97), peritoneal signs (10.58, 4.23e26.4),
CT
and BUN >25 mg/dL (3.08, 1.54e6.17), whereas presence of both PN and PVG (versus only
Ischemia
one) was associated with an increase (but not statistically significant increase) in odds
Necrosis
(2.01, 0.94e4.36). Although several ad hoc models were used to maximize diagnostic ability, with maximal odds ratio ¼ 174, the areas of receiver operating characteristic curves were all below 0.80, revealing suboptimal accuracy to diagnose I/N. Conclusions: Older age, peritoneal signs, and high BUN are associated with I/N, suggesting an ability to predict which patients need operation. CT findings traditionally suggestive of ischemic PN/PVG, however, do not diagnose I/N accurately enough to reliably identify patients needing operation. ª 2013 Elsevier Inc. All rights reserved.
* Corresponding author. Department of Surgery, Saint Agnes Hospital Center, University of Maryland, 900 Caton Avenue, Mailbox #207, Baltimore, MD 21229. Tel.: þ1 410 368 2748; fax: þ1 410 951 4007. E-mail address:
[email protected] (S.C. Cunningham). 0022-4804/$ e see front matter ª 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jss.2013.06.006
582
1.
j o u r n a l o f s u r g i c a l r e s e a r c h 1 8 5 ( 2 0 1 3 ) 5 8 1 e5 8 6
Introduction
Pneumatosis (PN) and portal venous gas (PVG) can be ominous radiographic signs, whose clinical significance is not well understood. PN is the abnormal accumulation of gas within the bowel wall and PVG is the presence of gas within the portomesenteric venous system, typically within the liver, where it tends to be identified in the periphery of the liver, in contradistinction to the more centrally located air of pneumobilia. PVG, and especially the more common PN, may be associated with a spectrum of illnesses ranging from grave illness due to intestinal ischemia or necrosis (I/N) to a physiologically well patient with an incidental radiographic finding of little clinical consequence. Unfortunately, the factors that aid clinicians in determining where on this spectrum a given patient lies are not well described, owing to the rarity of these radiographic findings and the paucity of data from singleinstitution studies with small data sets that currently exist in the literature. To further elucidate whether different types of PN and PVG or other factors (demographic, clinical, laboratory, or radiologic) can help clinicians in identifying patients with intestinal I/N, we have performed a retrospective multicenter study and constructed several scoring systems to guide clinicians in decision making regarding these patients.
2.
Methods
2.1.
Study population
After institutional review board approval from all four participating institutions (University of Maryland Medical Center, Baltimore Veterans Affairs Medical Center, Johns Hopkins Hospital, and the University of Michigan Hospital), a database was designed and populated with the demographic, clinical, radiographic, and laboratory data of patients who were identified to have either PN or PVG. These patients were identified by electronic searches of radiographic databases or manual searches of film libraries, or both, depending on the institution. Inclusion criteria were age 18 y and a computed tomography (CT) diagnosis of PN or PVG from 1983e2007. Exclusion criteria were age median [ 23) Arterial pH (% > median ¼ 7.39) Serum lactate (% > median ¼ 2.4) Serum albumin (% > median [ 2.6) Serum BUN (% > median [ 25) White blood count/1000 (% > median ¼ 11) Pneumoperitoneum (% yes) PVG and PN (% having both)y Bowel involvement Small Large Other Stranding (% yes) Thickening (% yes) Pneumatosis type (% linear or curvilinear versus cystic or bubbly) Ascites (% yes) Hypertension (% yes) Diabetes mellitus (% yes) Cardiovascular disease (% yes) Cerebrovascular accidents Immunosuppression
n*
209 194 193 203 87 88 163 203 198 202 209 203
199 200 177 206 205 205 205 200 195
Ischemia and/or necrosis
Unadjusted OR (95% CI)
Had I/N
Did not have I/N
61% 25% 42% 43% 51% 57% 44% 66% 57% 29% 34%
34% 18% 7% 62% 57% 36% 60% 37% 44% 20% 17%
3.00 (1.70e5.30) 1.57 (0.78e3.13) 9.40 (4.04e21.9) 0.47 (0.27e0.82) 0.78 (0.31e1.92) 2.38 (1.00e5.65) 0.52 (0.28e0.96) 3.21 (1.80e5.71) 1.70 (0.97e2.98) 1.66 (0.87e3.18) 2.52 (1.32e4.80)
50% 35% 15% 68% 61% 68% 64% 58% 16% 53% 22% 15%
42% 46% 12% 55% 53% 72% 48% 39% 21% 27% 7% 30%
1.00 (referent) 0.64 (0.35e1.17) 1.10 (0.47e2.60) 1.77 (0.99e3.16) 1.41 (0.80e2.48) 0.81 (0.42e1.55) 1.92 (1.09e3.36) 2.22 (1.26e3.89) 0.72 (0.35e1.48) 3.07 (1.71e5.51) 3.76 (1.56e9.06) 0.43 (0.21e0.89)
The bold values are statistically significant. * Number of patients for whom the results were available and who were included in the analysis. y PVG: portal venous gas. Since all patients enrolled in this study had either PVG or PN, this row compared those with both signs versus one.
celiac sprue [7]; small-bowel obstruction [8]; pyloric stenosis [9]; gastric ulcer [5]; graft-versus-host disease [10]; pseudoobstruction [11]; scleroderma [12]; diverticulitis [13]; cystic fibrosis [14]; acquired immunodeficiency syndrome [15]; chronic obstructive pulmonary diseases [16]; asthma [17]; a variety of medications, especially high-dose corticosteroids [18e21]; recent endoscopy [22]; and jejunal feeding tubes [23]. The clinical spectrum of the acuity of the many illnesses associated with PN or PVG is very broad, ranging from insignificant and incidental to severely life-threatening, and this breadth poses a dilemma to the surgeon, who may have little concrete guidance in determining where on this spectrum a given patient lies. This lack of reliable guidance is due in large part to the rarity of PN and PVG, and to the singleinstitution nature of the small studies that currently exist in the literature. Other studies, smaller and single-institution, in contradistinction to the present study, have also evaluated similar clinical, radiologic, and laboratory parameters to differentiate between cases with I/N versus other diagnoses. Greenstein et al. [24], for instance, found that a history of emesis, age >60 y, leukocytosis, PVG, sepsis, and acidosis were associated with need for laparotomy. Wayne et al. [25] developed a vascular risk factor score, including history of smoking, diabetes, hypertension, hyperlipidemia, coronary artery disease, peripheral vascular disease, and vasculitis, as well as an abnormal abdominal physical exam and lactic acidosis in the setting of small-bowel PN. The findings of Wayne and Greenstein differed from each other, and the findings of the present study were different
from both. These differences could be due to differences in sample size, differences in choice of variables, the retrospective design, or the multicenter nature of our study. In comparison to the study of Greenstein et al., the current study found age >60, acidosis, and PVG (in addition to PN) to be significantly associated with intestinal I/N in univariate analyses, but leukocytosis was not, and none of these parameters was significantly associated in multivariable analyses except age 60 y, peritoneal signs on physical exam, and BUN >25 mg/dL. Wayne et al. developed their model based on only 74 patients, with an additional 14 patients for their validation set. The current study did not use exactly the same variables as Wayne et al., as lactate data were available for only 89 patients. Although data on peripheral vascular disease were not included here, many other variables were collected beyond those of Wayne et al. In the present series, like that of Wayne et al., abdominal pain (tenderness associated with peritonitis on physical exam) was associated with higher risk of I/N, but several other variables used in that model, including history of coronary artery disease, hypertension, and hyperlipidemia, were associated with risk of outcome only in univariable models, but not in multiple regression models. This change in significance might be due to the fact that several of the variables used in the model were correlated; for example, age is significantly associated with diabetes and cardiovascular disease, and after its inclusion in the model, other variables lost significance. Given the rarity of PN and PVG, and the resultant lack of prospective data, surgeons and radiologists have relied on often-anecdotal observations, such as CT characteristics
585
j o u r n a l o f s u r g i c a l r e s e a r c h 1 8 5 ( 2 0 1 3 ) 5 8 1 e5 8 6
Table 2 e Association of variables with the outcome in multivariable logistic regression models. n ¼ 137
Variable
n ¼ 166
Adjusted OR (95% CI)*; P value Age (% 60 y) Peritoneal signs (% yes) Arterial CO2 (% > median ¼ 23) Serum albumin (% > median ¼ 2.6) Serum BUN (% > median ¼ 25) PVG and PN (% having both) Stranding (% yes) Ascites (% yes) Hypertension (% yes) Cardiovascular disease (% yes) Cerebrovascular accidents Immunosuppression
2.84 13.4 0.59 0.55 3.85 2.40 0.85 1.63 0.97 1.65 1.25 0.46
(1.07e7.52); 0.03 (4.22e42.7); 25 mg/dL were significantly associated with I/N. However, none of the models using these data could diagnose I/N with an accuracy that was good enough for clinical purposes. To increase the possibility of finding such models, we attempted several models and ad hoc analyses. These included using stepwise forward and backward logistic regression models with a P value cutoff point of 0.10. However, even using such models, sensitivity ranged from 0.51e0.73 and specificity
ranged from only 0.69e0.90, with all models having areas under the ROC curves below 0.80. A truly useful clinical test, by contrast, should have a much higher accuracy. Our analysis study included an evaluation set but did not use a validation set of patients. This would have been a problem if results had suggested that the scoring system could diagnose perforation or necrosis accurately. However, despite analyzing data in several ways and building multiple ad hoc models, results were mostly negative; therefore, we did not require a validation set. There are a number of important limitations to this study. Because this is a retrospective study, it is prone to several biases, not all of which are possible to control for. Because of its multicenter nature, technical limitations prevented interstate sharing of images, resulting in the unfortunate necessity that different radiologists (albeit all subspecialists) interpreted the CT images from different institutions. Although multicenter studies have greater statistical power and broader applicability, they may have the disadvantage of higher
Fig. e CT images from two patients with PN. (A) True-positive case. CT scan with intravenous contrast. Patient with pneumatosis and surgically proven ischemia of both the stomach and duodenum. Note the intramural gas and thickened nonenhancing stomach and duodenal wall (solid arrows) compared with the normally enhancing thin wall of an adjacent distal small bowel loop (interrupted arrow). (B) False-positive case. CT scan without intravenous contrast. Patient with abdominal pain and extensive pneumatosis of small bowel loops (solid arrows). Surgical exploration revealed no ischemia. The pneumatosis was considered secondary to the recent insertion of a percuateous jejunostomy tube (interrupted arrow).
586
j o u r n a l o f s u r g i c a l r e s e a r c h 1 8 5 ( 2 0 1 3 ) 5 8 1 e5 8 6
background noise, thereby increasing misclassification and biasing the odds ratios toward null. Although retrospective multicenter studies may lack a uniform mode of data collection, their results are closer to the daily practice of physicians. Finally, due to the very long (24-y) period of the study, differences in imaging technology are not accounted for.
5.
Conclusion
PN and PVG are uncommon CT findings known to be associated with a wide range of clinical scenarios, ranging from incidental and insignificant to severely life-threatening. Although the scoring system developed in this study may be a helpful guide in some patients, as odds ratios over 100 were detected for the highest versus lowest scores, there was no single variable or single cutoff point using multiple variables that could accurately diagnose I/N. Therefore, the greater value of this study may well be in demonstrating, in a large number of patients, that CT findings thought to be reassuring, such as the bubbly or cystic type of PN and the lack of bowelwall thickening and adjacent mesenteric stranding, should not reassure the examining surgeon.
Acknowledgment The authors thank Dr Eike Gallmeier for translation of and assistance with the early reports of PN in German and Latin, Dr Kimberly Lumpkins for analysis and interpretation of data, and Dr Hwa Yeon Lee for acquisition of data.
Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jss. 2013.06.006.
references
[1] Duvernoi JG. Aer intestinorum tam sub extima quam intima tunica inclussus: observationes anatomicae. Acad Acient Imp Petropol 1730;5:213. [2] Koss LG. Abdominal gas cysts (pneumatosis cystoides intestinorum hominis); an analysis with a report of a case and a critical review of the literature. AMA Arch Pathol 1952;53:523. [3] Wiesner W, Mortele KJ, Glickman JN, Ji H, Ros PR. Pneumatosis intestinalis and portomesenteric venous gas in intestinal ischemia: correlation of CT findings with severity of ischemia and clinical outcome. AJR Am J Roentgenol 2001;177: 1319. [4] Kernagis LY, Levine MS, Jacobs JE. Pneumatosis intestinalis in patients with ischemia: correlation of CT findings with viability of the bowel. AJR Am J Roentgenol 2003;180:733. [5] Matsumoto A, Isomoto H, Shikuwa S, et al. Pneumatosis intestinalis in ulcerative colitis. Med Sci Monit 2009;15:CS139.
[6] John A, Dickey K, Fenwick J, Sussman B, Beeken W. Pneumatosis intestinalis in patients with Crohn’s disease. Dig Dis Sci 1992;37:813. [7] Nathan H, Singhal S, Cameron JL. Benign pneumatosis intestinalis in the setting of celiac disease. J Gastrointest Surg 2006;10:890. [8] Alkhatib AA, Elkhatib FA, Alkhatib OF, Zurcher R. Pneumatosis intestinalis and gas in portal vein associated with small bowel obstruction. J Emerg Med 2011;40:e125. [9] Gutierrez Andreu M, Otero Perpina B, Munoz C, De La Cruz Vigo F. [Gastric pneumatosis secondary to a pyloric stenosis of peptic origin]. Cir Esp 2011;89:188. [10] Schulenburg A, Herold C, Eisenhuber E, et al. Pneumatosis [correction of Pneumocystis] cystoides intestinalis with pneumoperitoneum and pneumoretroperitoneum in a patient with extensive chronic graft-versus-host disease. Bone Marrow Transplant 1999;24:331. [11] Le P, Benazzouz A, Fritsch L. [A pneumatosis cystoides coli revealed by a pseudo-obstruction]. Ann Chir 2003;128:117. [12] Seket B, Kaczmarek D, Tiffet O. Intestinal pseudoobstruction and pneumatosis cystoides intestinalis in a scleroderma patient. J Am Coll Surg 2007;205:180. [13] Sammon AM, Malcolm AJ. Pneumatosis cystoides intestinalis secondary to chronic jejunal diverticulitis. Postgrad Med J 1979;55:820. [14] Hernanz-Schulman M, Kirkpatrick J Jr, Shwachman H, Herman T, Schulman G, Vawter GF. Pneumatosis intestinalis in cystic fibrosis. Radiology 1986;160:497. [15] Op de Beeck B, Peters K, Spinhoven MJ, Snoeckx A, Salgado R, Parizel PM. Asymptomatic pneumatosis intestinalis in AIDS. JBR-BTR 2009;92:253. [16] Doumit M, Saloojee N, Seppala R. Pneumatosis intestinalis in a patient with chronic bronchiectasis. Can J Gastroenterol 2008;22:847. [17] Culver GJ. Pneumatosis intestinalis with associated retroperitoneal air. Report of a complication of severe asthma. JAMA 1963;186:160. [18] Vogel Y, Buchner NJ, Szpakowski M, Tannapfel A, Henning BF. Pneumatosis cystoides intestinalis of the ascending colon related to acarbose treatment: a case report. J Med Case Rep 2009;3:9216. [19] Goodman RA, Riley TR 3rd. Lactulose-induced pneumatosis intestinalis and pneumoperitoneum. Dig Dis Sci 2001;46: 2549. [20] Kopp AF, Gronewaller E, Laniado M. Pneumatosis cystoides intestinalis with pneumoperitoneum and pneumoretroperitoneum following chemotherapy. Abdom Imaging 1997;22:395. [21] Day DL, Ramsay NK, Letourneau JG. Pneumatosis intestinalis after bone marrow transplantation. AJR Am J Roentgenol 1988;151:85. [22] Cammarota G, Piscaglia AC. Free peritoneal gas accumulation caused by pneumatosis coli after diagnostic gastrointestinal endoscopy. Clin Gastroenterol Hepatol 2009;7:A18. [23] Lebiedz P, Ullerich H, Seelig M, Domschke W, Kucharzik T. Jejunal feeding tube causing intestinal pneumatosis and fulminant portal venous gas embolism. Endoscopy 2005; 37:926. [24] Greenstein AJ, Nguyen SQ , Berlin A, et al. Pneumatosis intestinalis in adults: management, surgical indications, and risk factors for mortality. J Gastrointest Surg 2007;11: 1268. [25] Wayne E, Ough M, Wu A, et al. Management algorithm for pneumatosis intestinalis and portal venous gas: treatment and outcome of 88 consecutive cases. J Gastrointest Surg 2010;14:437.
