Evidence for use of damage control surgery and damage control interventions in civilian trauma patients: a systematic review

Background Although damage control (DC) surgery is widely assumed to reduce mortality in critically injured patients, survivors often suffer substantial morbidity, suggesting that it should only be used when indicated. The purpose of this systematic review was to determine which indications for DC have evidence that they are reliable and/or valid (and therefore in which clinical situations evidence supports use of DC or that DC improves outcomes). Methods We searched 11 databases (1950–April 1, 2019) for studies that enrolled exclusively civilian trauma patients and reported data on the reliability (consistency of surgical decisions in a given clinical scenario) or content (surgeons would perform DC in that clinical scenario or the indication predicted use of DC in practice), construct (were associated with poor outcomes), or criterion (were associated with improved outcomes when DC was conducted instead of definitive surgery) validity for suggested indications for DC surgery or DC interventions. Results Among 34,979 citations identified, we included 36 cohort studies and three cross-sectional surveys in the systematic review. Of the 59 unique indications for DC identified, 10 had evidence of content validity [e.g., a major abdominal vascular injury or a packed red blood cell (PRBC) volume exceeding the critical administration threshold], nine had evidence of construct validity (e.g., unstable patients with combined abdominal vascular and pancreas gunshot injuries or an iliac vessel injury and intraoperative acidosis), and six had evidence of criterion validity (e.g., penetrating trauma patients requiring > 10 U PRBCs with an abdominal vascular and multiple abdominal visceral injuries or intraoperative hypothermia, acidosis, or coagulopathy). No studies evaluated the reliability of indications. Conclusions Few indications for DC surgery or DC interventions have evidence supporting that they are reliable and/or valid. DC should be used with respect for the uncertainty regarding its effectiveness, and only in circumstances where definitive surgery cannot be entertained. Supplementary Information The online version contains supplementary material available at 10.1186/s13017-021-00352-5.


Background
In patients requiring operative intervention after major trauma, surgeons must decide whether to perform a definitive or damage control (DC) procedure [1,2]. As opposed to definitive surgery (where all injuries requiring repair are repaired and the explored cavity closed), DC surgery involves quickly controlling exsanguinating hemorrhage and/or gross contamination using one or more abbreviated (or DC) interventions [2]. The patient is subsequently admitted to the intensive care unit (ICU) for ongoing resuscitation with the goal of restoring preinjury physiology before returning to the operating room for additional surgery [2][3][4].
Although widely assumed to reduce mortality in critically injured patients [5], survivors of DC surgery have been reported to have a high risk of complications (e.g., intra-abdominal sepsis, enteric fistulae, and complex ventral herniae) and often suffer long lengths of ICU and hospital stay [2,4,[6][7][8][9][10][11]. It is therefore important to ensure that DC surgery is only performed on patients in which the expected survival benefit of the procedure outweighs its expected risk of negative consequences [1]. Despite this, the benefit/risk profile of using DC surgery in different clinical situations has not been comprehensively evaluated, and several authors have recently reported data suggesting that substantial variation in use of DC surgery exists across trauma centers or that it may be overused [12][13][14][15].
We hypothesize that variation in use of DC surgery may be at least partially explained by the uncertainty that exists as to when the procedure is indicated [1,2]. Thus, we recently initiated a program of research to develop evidence-informed indications for the appropriate use of DC surgery and DC interventions in civilian trauma patients [1][2][3][16][17][18]. The purpose of this systematic review and meta-analysis was to determine which indications for DC surgery and DC interventions in civilian trauma patients have evidence supporting that they are reliable and/or valid (and in which clinical situations evidence supports use of DC or that DC improves outcomes). The data reported in this study therefore provide a comprehensive assessment of the reported studies evaluating whether use of DC instead of definitive surgery is associated with improved outcomes in injured patients.

Protocol
Study methods were pre-specified in a protocol developed according to the Preferred Reporting Items in Systematic Reviews and Meta-analyses [19] and Metaanalysis of Observational Studies in Epidemiology [20] statements.

Search strategy
Using published search strategies designed for identifying indications for DC surgery and DC interventions in trauma patients, we searched MEDLINE, EMBASE, PubMed, Scopus, Web of Science, and the Cochrane Library from their inception to April 1, 2019 without restrictions (see Supplementary Table 1 in our published protocol paper [1] for details of our electronic bibliographic database search strategies). We also used the PubMed "related articles" feature and searched references from included and relevant review articles and abstracts from conferences held between 2009 and 2015, including meetings of the American Association for the Surgery of Trauma (AAST), Australasian Trauma Society, Eastern Association for the Surgery of Trauma (EAST), Trauma Association of Canada, and Western Trauma Association (WTA). To identify unpublished studies, we searched 12 trauma organizational websites [1] and Google Scholar (the first 10 web pages) using combinations of the terms trauma, injury, abbreviated surgery, bailout surgery, damage control, damage control surgery, indication, and predictor.

Study selection
Two investigators (D.J.R., N.B.) independently screened the titles and abstracts of citations identified by the search and selected articles for full-text review. We included full-text studies that reported original data on the reliability or validity of suggested indications for DC surgery or DC interventions in civilian trauma patients. We excluded studies that included only patients injured in combat or by thermal mechanisms or focused exclusively on DC for emergency general or vascular surgery or orthopedic or neurologic injuries. Study eligibility disagreements were resolved by consensus between the two investigators.

Definitions
An indication was defined as a clinical finding/circumstance or scenario that reportedly advised use of DC surgery (or a DC intervention) over definitive surgery (or a definitive surgical intervention) [1]. DC surgery was broadly defined as a multi-step operative intervention, which included an abbreviated initial surgical procedure that aimed to rapidly control bleeding and/or gross contamination [1]. We did not predefine DC interventions. Instead, we included articles that satisfied the above criteria where an indication was reported for a surgical intervention suggested by authors to constitute DC or an abbreviated surgical technique [e.g., temporary abdominal closure (TAC)/open abdominal management after trauma laparotomy] [1,2].
Indication reliability was defined as the degree to which the same decision to conduct DC was made when surgeons were provided the same clinical finding/scenario (test-retest) or when encountered by the same surgeon (intra-rater) or different surgeons (inter-rater). Validity included content, construct, and criterion validity (see Table 1 for detailed definitions of these measures) [21,22]. Content validity was defined as the extent to which surgeons reported that they would perform DC in a given clinical scenario or that an indication predicted use of DC in practice [21,22]. Construct validity referred to how well one indication or a combination of indications and demographic variables predicted poor outcomes in patients not treated with DC (i.e., the extent to which an indication was associated with a higher probability of poor outcomes in patients treated with definitive surgery and therefore should be considered as a potential indication for DC) [21,22]. Criterion validity referred to the extent to which the utilization or conduct of DC instead of definitive surgery for one or more indications was associated with improved patient outcomes [21,22].

Data extraction
Two investigators (D.J.R., N.B.) independently extracted data from included studies into pilot-tested tables summarizing characteristics of the included studies and the content, construct, and criterion validity of suggested indications for DC surgery and DC interventions (Table  1). An interpreter assisted with data extraction for one Russian [23] and two Mandarin Chinese [24,25] language studies. We extracted data on (1) study design, setting, and participants; (2) suggested indications for DC surgery or DC interventions as reported by study authors; and (3) measures of indication reliability and validity. For content validity, we extracted data on the percentage of surgeons that reported that they would perform DC in a given clinical scenario or odds ratios (ORs) or hazard ratios (HRs) [with surrounding 95% confidence intervals (CIs)] indicating the degree to which that indication predicted use of DC in practice. For construct validity, we extracted data on how well one indication or a combination of indications predicted outcomes in patients not treated with DC. Finally, for criterion validity, we extracted data on the extent to which the utilization or conduct of DC instead of definitive surgery for one or more indications was associated with patient outcomes. Outcomes of interest for the assessment of construct and criterion validity included survival, development of coagulopathy, reported measures of morbidity, and lengths of hospital and ICU stay. Outcomes were extracted at the longest follow-up duration. When both unadjusted and adjusted outcome estimates were reported, the most adjusted estimate was extracted.

Risk of bias assessment
The same two investigators independently evaluated study risk of bias. Cohort studies were assessed using an expanded version of the Quality in Prognosis Studies How well one indication or a combination of indications and demographic variables predicted poor outcomes in patients not treated with DC (i.e., the extent to which an indication was associated with a higher probability of poor outcomes in patients treated with definitive surgery and therefore should be considered as a potential indication for DC) In a cohort study, the intraoperative identification of a major abdominal vascular injury was associated with an increased risk of mortality in patients who underwent definitive laparotomy for trauma Criterion Extent to which the indication related to a reference standard (e.g., the extent to which conducting DC instead of definitive surgery in that clinical situation was associated with an improvement in outcomes) Extent to which the utilization or conduct of DC instead of definitive surgery for one or more indications was associated with improved patient outcomes In a cohort study, use of DC instead of definitive surgery for patients with a major abdominal vascular injury was associated with an improvement in in-hospital adjusted mortality Where CI indicates confidence interval; DC, damage control; and OR, odds ratio tool [26,27], which included questions regarding study participation and attrition; indication or outcome description and measurement; confounding measurement and account; whether the operative profile chosen (i.e., DC versus definitive surgery) may have varied in relation to the indication of interest; and methods and reporting of statistical analyses (see Supplemental Digital Content 1 for the operationalized list of quality domains evaluated) [26][27][28]. For cross-sectional studies, we evaluated sampling methods, response rates, and whether the reported methods would permit replication; sample was representative of the population; questionnaire was adequately described, pretested, and had evidence of reliability and/or validity; statistical methods; and if all respondents were accounted for [29]. The assessment of statistical analyses incorporated recommendations for appraising logistic regression models [30,31]. Disagreements regarding risk of bias assessments were resolved by consensus.

Data synthesis
We used directed qualitative content analysis to group unique indications into the subcategories and categories of a previously developed framework for conceptualizing indications for DC [3,32]. We then used a vote counting scale [33] to incorporate our risk of bias assessments into the synthesis of evidence regarding whether indications were reliable and/or valid [26]. The aggregate scale summarized strength of evidence as (1) not reported, (2) inconclusive (no evidence or a low to moderate association in the setting of an overall high amount of bias in at least one quality domain), (3) a consistently strong association in the setting of an overall high amount of bias in only one quality domain, (4) a consistently low to moderate association in the setting of an overall moderate amount of bias in one or more quality domains, (5) a consistently low to moderate association with a low amount of bias in all quality domains or a consistently strong association with an overall moderate amount of bias in one or more study quality domains, and (6) a consistently strong association with a low overall amount of bias in all study domains.

Statistical analyses
Inter-investigator agreement regarding full-text article inclusion was quantified using kappa (κ) statistics [34]. We summarized dichotomous data using counts (percentages) and compared them using ORs with 95% CIs or Fisher's exact tests. We combined adjusted ORs for indications with similar definitions using Mantel-Haenszel-weighted DerSimonian and Laird randomeffects models [35]. Heterogeneity in these estimates were assessed using I 2 statistics and tests of homogeneity [36,37]. We considered two-sided p values < 0.05 statistically significant. Stata MP version 13.1 (Stata Corp., College Station, TX) was used for statistical analyses.

Description of included studies
Characteristics of included cohort and cross-sectional studies are presented in Table 2 [69]) and one of Trauma Association of Canada surgeon members (administered in 2004; results of which were reported across two publications) [70,71]. Surgeon response rates varied from 26% [69] to 84% [70,71]. Of the respondents, two studies reported that 73-85% practiced in academic centers [69,72] and one that 45% specialized in trauma and critical care [69].

Risk of bias assessment
The risk of bias assessment for the 36 included cohort studies is outlined in Supplemental Digital File 3. Nine (25.0%) studies were at high risk of study participation bias, 17 (47.2%) reported outcome comparisons at high risk of confounding, and 28 (77.8%) did not report follow-up duration. Six of the nine studies at high risk of study participation bias evaluated the construct validity of indications among patients treated only with DC surgery (and therefore outcomes associated with these variables are better interpreted as predictors of poor outcome after DC surgery rather than measures of indication construct validity) [47,48,56,61,63,66]. Of the 17 cohort studies that adjusted estimates using logistic regression, these analyses were at a moderate risk of bias in 12 studies and a high risk of bias in one study, largely because of a low or unclear number of events per variable and inadequate reporting of methods used to select predictor variables and/or build models (Supplemental Digital Content 4).
The risk of bias assessment for the three included cross-sectional studies is included in Supplemental Digital File 5. Two studies described methods that would permit replication, identified a sample that was likely representative of the broader trauma surgical community, and adequately described the questionnaire (or provided it as supplementary material) [70][71][72]; however, only one reported that the questionnaire was pretested [70,71] and none provided evidence of questionnaire reliability or validity.

Content validity
The cross-sectional studies [69][70][71][72] evaluated the content validity of indications for TAC after trauma laparotomy by asking surgeons whether open abdominal management was indicated in different clinical situations (see Table 3 for assessments of content validity in crosssectional and cohort studies). Most respondents supported use of TAC when they were unable to close the fascia (or closure was subjectively tight), there was massive visceral edema, planned reoperation was required (e.g., to remove packs or perform a "re-look"      [69][70][71][72]. Of the cohort studies that evaluated indication content validity (i.e., whether the indication predicted use of DC in practice), Hirshberg et al. reported that an artificial neural network containing torso bullet wound location/trajectory pattern and systolic blood pressure (BP) in the emergency department (ED) had a high sensitivity (83%) and specificity (93%) for identifying patients treated with DC laparotomy [55]. Further, Watson et al. reported that a major abdominal vascular injury was independently associated with the decision to conduct DC laparotomy among patients enrolled in in the Pragmatic, Randomized Optimal Platelet and Plasma Ratios (PROPPR) randomized trial [15]. Another study by Leppäniemi et al. reported that all patients with a AAST grade III-V liver and IV-V spleen injury underwent DC laparotomy while 42% of those without this injury pattern did not (p = 0.02) [46]. Finally, Savage et al. reported that packed red blood cell (PRBC) transfusion volumes exceeding the critical administration threshold (≥ 3 U PRBCs in 1 hour of the first 24 h of injury) were independently associated with a dosedependent increase in use of DC laparotomy [39].

Construct validity
The construct validity of indications (or predictive models containing a combination of indications, other clinical findings, and/or baseline demographic variables) for DC surgery was evaluated in 23 studies, which examined associations between indications and survival or coagulopathy (Table 4). Three studies reported that a high ISS score, preoperative hypothermia, an elevated base deficit in the pre-or intraoperative setting, and the identification of a combined pancreas and abdominal vascular injury during operation were independently associated with decreased survival in patients mostly treated with definitive surgery, suggesting that DC should be considered in these high-risk scenarios [44,57,64]. Interestingly, however, a preoperative pH < 7.20 was also independently associated with decreased survival among injured patients who received DC laparotomy in another study (suggesting that it was also a poor prognostic factor among those treated with DC) [47]. Two other studies evaluated the association between development of a laboratory-confirmed coagulopathy (variably defined), a clinical scenario where DC has long been recommended over definitive surgery, and an ISS score > 25, systolic BP < 70 mmHg, or lowest    Variables reported to be entered into the regression model included those associated with mortality (p < 0.20) that did not have > 10% missing data. These may have included, at a minimum, systolic BP and respiratory rate in the ED; Glasgow Coma Scale score, ISS, and Revised Trauma Scale score; preoperative hematocrit; crystalloids and blood given in the ED; estimated intraoperative blood loss; crystalloids and blood given in the OR; total fluids; and length of stay in the surgical intensive care unit and hospital b Variables reported to be entered into the logistic regression model included those associated with survival in bivariate analysis (p < 0.20). These appeared to at least include ISS > 20; RTS > 0; GCS ≤ 3 or < 9; MOI; absence of spontaneous ventilation, a palpable carotid pulse, or extremity movement; non-sinus rhythm on the electrocardiogram; systolic BP and respiratory rate as a 3-level or 2-level categorical variable; a pulmonary artery and vein, thorax, thoracic or abdominal vascular, or liver injury; thoracotomy or laparotomy in the OR; coagulopathy; dysrhythmia; and type of dysrhythmia c Defined by the authors as the perceived need to initiate DC maneuvers by a surgical attending, which was reported to be subjective, but usually occurred in the setting of major blood loss, hypothermia, acidosis, and the presence of multiple injuries [53] d Defined by the authors as diffuse bleeding from all wounds without discrete bleeding vessels, absence of observable clots, prolonged PT and PTT along with decreased platelet count, or decreased platelet count alone [66] e In this study, the probability of developing coagulopathy (defined as a PT and PTT > 2 times that of normal laboratory control) in patients who had received a transfusion of > 10 Us PRBCs in the first 24 h was 10% for those with an ISS > 25; 39% for those with an ISS > 25 and lowest systolic BP < 70 mmHg; 58% for those with an ISS > 25 and lowest pH < 7.1; 49% for those with an ISS > 25 and lowest temperature < 34°C; 85% for those with an ISS > 25 and lowest systolic BP < 70 mmHg and temperature < 34°C; and 98% for those with an ISS > 25 and lowest systolic BP < 70 mmHg, pH < 7.1, and temperature < 34°C. f Variables entered into the logistic regression model included FFP:PRBC ratio at 6 h; age > 55 years; ISS > 25; PRBC, FFP, and platelet U transfused at 6 h; crystalloids in 24 h; and ED systolic BP < 70 mmHg, temperature < 34°C, and pH < 7.1 temperature < 34°C or pH < 7.1 in trauma patients transfused > 10 units of PRBCs in the first 6 or 24 h [50,60]. Among these studies, the pooled adjusted OR for development of a laboratory-confirmed coagulopathy among patients with an ISS score > 25 was 6.11 (95% CI, 1.68-22.16; I 2 = 0%; heterogeneity p = 0.65), systolic BP < 70 mmHg was 1.66 (95% CI, 0.15-19.10; I 2 = 79.5%; heterogeneity p = 0.03), lowest temperature < 34°C was 7.12 (95% CI, 2.53-20.05; I 2 = 0%; heterogeneity p = 0.74), and lowest pH < 7.1 was 4.14 (95% CI, 0.60-28.67; I 2 = 74.2%; heterogeneity p = 0.05).

Criterion validity
Two studies evaluated outcomes associated with implementation or utilization of indications for DC surgery while 14 compared outcomes of patients treated with DC versus definitive surgery in different clinical situations (Table 5). Rice et al. reported that, when compared to only minor deviations, moderate or major deviations from a protocol that suggested use of DC surgery in injured patients with a temperature < 35°C, lactate > 4 mmol/L (or more than twice the upper limit of normal), or corrected pH < 7.3 was independently associated with reduced survival [43]. Asensio et al. reported that implementing a guideline that suggested use of DC surgery for trauma patients with one of 12 different clinical findings/events was associated with a decreased unadjusted odds of infections, an increased unadjusted odds of abdominal wall closure, and reduced unadjusted lengths of ICU and hospital stay [52].
Of the remaining 14 studies, Chinnery et al., Rotondo et al., and Stone et al. observed a large improvement in unadjusted survival when DC or staged laparotomy was used instead of definitive surgery to manage unstable patients with combined abdominal vascular and pancreas gunshot injuries, who received > 10 U PRBCs and had ≥ 1 major abdominal vascular and ≥ 2 abdominal visceral injuries, or that developed a coagulopathy during operation, respectively [44,62,68]. In contrast, Harvin et al. reported that after matching injured patients on propensity scores created using 17 different variables, use of DC instead of definitive laparotomy (for intra-abdominal packing, a second-look laparotomy, hemodynamic instability, to expedite postoperative care or intervention, or for other reasons) was associated with a significantly increased incidence of gastrointestinal (GI) ileus, GI bleeding, abdominal fascial dehiscence, superficial surgical site infection, and death [38]. Moreover, Martin et al. reported that use of DC laparotomy in trauma patients with an arrival systolic BP > 90 mmHg, no severe TBI, and no combined abdominal injuries was associated with an increased adjusted odds of major postoperative complications and an increased adjusted length of hospital stay when compared to patients with a severe abdominal injury who underwent therapeutic definitive laparotomy [14].

Narrative synthesis of validity of indications for use of DC surgery
The narrative synthesis of the aggregate evidence for use of indications for DC surgery is presented in Table 6. Of the 59 unique indications identified using directed qualitative content analysis, two had moderate or strong evidence of content validity [upper quadrant abdominal gunshot wound with a horizontal shift trajectory (e.g., from the right to the left upper quadrant) and a systolic BP < 105 mmHg or right upper quadrant wound with a bullet retained in the same quadrant and a systolic BP < 90 mmHg). Further, nine had moderate or strong evidence of construct validity (high ISS score, preoperative hypothermia, unstable patients with combined abdominal vascular and pancreas gunshot injuries, and transfusion > 10 U PRBCs and ISS score > 25 or lowest temperature < 34°C in the pre-or intraoperative setting) and six had moderate or strong evidence of criterion validity (pre-or intraoperative hypothermia, increased lactate, or decreased pH).

Discussion
This systematic review is the first to comprehensively compile and critically analyze the evidence for use of DC surgery and DC interventions in civilian trauma patients. We identified 39 studies that evaluated the content, construct, and/or criterion validity of 116 indications for DC surgery. Most were single-center cohort studies that recruited relatively small samples of critically injured patients. Of the 59 unique indications identified using content analysis, 10 had evidence of content validity (i.e., surgeons self-reported that they would perform DC in that clinical scenario or the indication predicted use of DC in practice), nine had evidence of construct validity (i.e., were associated with poor outcomes in patients largely treated with definitive surgery, suggesting that DC be considered in these high-risk scenarios), and six had evidence of criterion validity (may be associated with improved outcomes when utilized or when DC was performed instead of definitive surgery).
Most included studies attempted to validate indications for use of DC surgery by assessing if they were associated with poor outcomes (i.e., coagulopathy or reduced survival); however, few studies subsequently sought to determine if DC improves survival in these situations and nearly one quarter of these studies included patients who only underwent DC (and therefore outcomes associated with these variables are better interpreted as predictors of poor outcome after DC surgery than measures of indication construct validity). Further, some physiology-based indications were associated with   [47]. This last finding suggests that although several measures of physiological compromise have been assessed by both international experts and practicing surgeons to be highly appropriate indications for use of DC [3,16,17], some data suggest that there may be a point beyond which physiologic derangements have progressed too far for DC surgery to improve survival [49]. We identified only six indications that had evidence to support that their utilization or the conduct of DC surgery may improve patient survival. These indications represent those with the most evidence to support their use and include the finding of hypothermia or acidosis, development of a coagulopathy during operation, or the identification of two injury patterns that preclude expedient definitive repair (combined abdominal vascular and pancreas gunshot injuries and ≥ 1 major abdominal vascular and ≥ 2 abdominal visceral injuries in patients who have received > 10 U PRBCs) [44,62,68]. However, as these were observational studies and operative profile (i.e., DC versus definitive surgery) was not randomly assigned, there were likely other, unmeasured reasons why surgeons chose to perform DC in these studies that are related to the risk of future outcomes (i.e., the studies were likely confounded by indication) [73].
Those indications with evidence suggesting that they may be associated with poor outcomes or that outcomes may be improved with use of DC surgery represent targets for focused future research efforts ( Table 6). As studies cannot deliver an unbiased and meaningful assessment of validity if the type of operative procedure varies in relation to the indication of interest [28], subsequent study of indications for DC surgery must compare the outcomes of performing DC versus definitive surgery in patients with well-defined indications. This should begin with prospective cohort studies designed to estimate the causal effects of DC surgery by controlling for confounding by indication using multivariate adjustment, propensity scores, or other techniques [73]. A randomized controlled trial would provide the least biased estimates of the benefit/harm ratio of DC compared to definitive surgery in different clinical circumstances. Initial trials should randomize patients to DC or definitive surgery in those clinical circumstances with the greatest uncertainty about the potential role of DC surgery (i.e., those indications listed in Table 6 that have no or equivalent evidence of content, construct, and criterion validity). While the above studies are being designed and conducted, creation of a list of DC consensus indications may allow for the conduct of quality improvement or knowledge translation interventions to decrease overutilization of DC in trauma patients. Where minor deviations included departures deemed not clinically significant; moderate deviations included care, which although departures were present, mostly followed protocol; and major deviations included those that did not meet the standards outlined in the protocol.
c Variables reported to be entered into the logistic regression model included age; gender; injury type; time from injury to hospitalization; PRBCs transfused before hospitalization; ISS; Glasgow Coma Scale score; shock; baseline hemoglobin, creatinine and activated PTT; country; and patients who did not require DC for both outcome comparisons d Variables reported to be entered into logistic and linear regression models included age; gender; mechanism of injury; head injury; major extremity injury; combined abdominal injury; ISS; presenting vitals; BD; and need for colon resection e In this study, 15 of the patients in the definitive laparotomy group were reported to ultimately need abdominal packing after conventional hepatic injury repair techniques. Moreover, 1 patient in the early therapeutic packing group received angiography before laparotomy