Introduction

Acute lower gastrointestinal bleeding (ALGIB) occurs with sudden hematochezia and is caused by various diseases, including colonic diverticular bleeding (CDB), colorectal cancer, and ischemic colitis. Although bleeding is often self-limiting, therapeutic interventions such as endoscopic hemostasis or transcatheter arterial embolization (TAE) are required for massive bleeding or rebleeding accompanied by hemorrhagic shock. If these interventions do not stop bleeding or if the hemorrhage cause is neoplastic, emergent or elective surgery is performed [1,2,3,4,5].

Although emergent surgery is generally considered as a high-risk intervention, there are a limited number of reports on surgery for ALGIB. Most of those included 10–60 surgery cases, and thus, data on implementation rate, indication, procedure, and mortality of the surgery remain insufficient to decide appropriate management. Surgery rates in ALGIB ranged from 0.2 to 46% [6,7,8,9,10,11,12,13,14,15,16,17,18] and mortality rates ranged from 5 to 27% depending on cohort or age of publication. A recent large-scale study reported that emergency colectomy performed in 1614 cases was associated with high 30-day mortality (12.2%) [11]. However, the dataset lacked diagnoses of ALGIB or bleeding sources, which are essential for proper management. Moreover, there are few reports on surgery-related risk factors and postoperative rebleeding. Because of these situations, evidence on indication, outcome, and risk factor of surgery for ALGIB remain limited and the guideline for ALGIB [19] expresses that the quality of the evidence regarding surgery for ALGIB is poor and mostly derived from small, retrospective reviews.

To identify factors leading to surgery and measures to avoid surgery as a severe outcome, we analyzed the data obtained by a multicenter large-scale cohort study of 10,342 acute hematochezia cases hospitalized from 2010 to 2019 by collecting more than 200 items of data, including surgery and its short- or long-term outcomes [20, 21].

Methods

Study design and patients

This retrospective multicenter cohort study, the CODE BLUE-J Study (COlonic DivErticular Bleeding Leaders Update Evidence from multicenter Japanese Study), was conducted at 49 hospitals across Japan. At each hospital, data were collected from the electronic medical records and entered into an Excel spreadsheet. This spreadsheet was sent to the secretariat’s institution to assess any omissions or errors in the data entered for each hospital, any data that needed to be corrected were emailed to the person responsible for the data at each hospital with comments on the Excel spreadsheet. The above process was repeated at least three times per hospital in order to obtain more complete data. A total of 10,342 adult patients were hospitalized for acute hematochezia between January 1, 2010 and December 31, 2019 [20, 21]. Of these patients, we retrospectively analyzed the detailed data of those who had undergone surgery. Approval was obtained from the ethical committees and institutional review boards of all 49 participating hospitals using the opt-out method. We excluded patients with insufficient clinical information and patient records of another hospitalization for ALGIB before admission in which surgery was performed (Fig. 1). To analyze surgery-related factors, we used the surgery group by excluding those who underwent surgery for adverse events induced by endoscopy or TAE and surgery for incidentally diagnosed lesions unrelated to ALGIB. In terms of emergency services, we classified the 49 participating hospitals into a high-volume group (more than 5000 emergency services in 2019) and a low-volume group (less than 5000 emergency services) according to our previous report [22]. In terms of hospital type, we classified these hospitals into an academic or non-academic and examined the relationship between surgery rates and the hospital-related factors.

Fig. 1
figure 1

Study Flow diagram of the study. *ALGIB acute lower gastrointestinal bleeding, **TAE transcatheter arterial embolization

Full size image

Variables and outcomes

We analyzed diseases indicated for surgery, intraoperative adverse events, postoperative adverse events (wound infection, anastomotic failure or bleeding, intestinal obstruction, intraperitoneal abscess, and wound bleeding), general postoperative adverse events (pneumonia, cardiovascular disorders, and others), the purpose of surgery, the urgency of surgery, preoperative identification of the bleeding site, postoperative rebleeding rate, number of days until postoperative rebleeding, observation period of postoperative rebleeding, and postoperative overall mortality rate. Rebleeding episodes were evaluated and defined as significant amounts of fresh bloody or wine-colored stool. According to the purposes, surgeries were classified into (i) surgery for hemostasis for severe refractory bleeding and (ii) surgery mainly not for hemostasis but resection of the original disease such as neoplasia. Furthermore, surgery for hemostasis was classified into (i) emergency surgery for severe refractory bleeding and (ii) elective surgery, in which bleeding has stopped once, but surgery is required for cancer or hemostasis due to the risk of rebleeding. Confidence levels for preoperative identification of the bleeding site were classified into (i) definitive, (ii) presumptive, or (iii) obscure. The definitive diagnosis was based on endoscopic visualization of SRH, which was determined stigmata of recent hemorrhage such as active bleeding, a visible vessel, or an adherent clot. The presumptive diagnosis was determined according to the presence of fresh blood localized in the colon, in which the patient had a potential bleeding source without SRH on colonoscopy, and a negative upper endoscopy, or small bowel endoscopy. The obscure diagnosis was determined according to no fresh blood in the colon, in which the patient had no potential bleeding source by colonoscopy with other tests showing negative results, including upper GI endoscopy or small bowel endoscopy. Of the 49 hospitals that participated in this study, 46 hospitals were able to perform small bowel capsule endoscopy or balloon-assisted endoscopy. Three hospitals were unable to perform small bowel examination except for CT scan. Surgical procedures were divided into (i) total or subtotal colectomy, (ii) segmental colectomy, (iii) segmental resection of the small bowel, and (iv) non-resection surgery. Moreover, the postoperative rebleeding rate, mortality rate, and observation period were examined as outcomes of hemostasis surgery.

In the analysis of risk factors associated with surgery, age < 65, sex, body mass index (BMI) > 25, current drinker, current smoker, performance status (PS) ≥ 3, syncope, fever, shock index (SI) > 1, diarrhea, abdominal pain, WBC > 10,000 (/μl), Hb < 7.0 (g/dl), Plt ≤ 15.0 (/μl), PT-INR ≥ 1.5, creatinine ≥ 1.5 (mg/dl), hemodialysis, albumin < 3.0 (g/dl), history of LGIB, Charlson comorbidity index (CCI) ≥ 2 [23], non-steroidal anti-inflammatory drugs (NSAIDs), low dose aspirin (LDA), non-LDA antiplatelet, anticoagulant, extravasation on CT, TAE, colonoscopy, stigmata of recent hemorrhage (SRH) detection, endoscopic hemostasis, transfusion need ≥ 6 units [24], in-hospital rebleeding, high-volume group, academic group, CDB, small bowel bleeding, colorectal cancer, hemorrhoids, inflammatory bowel disease, intestinal ischemia, rectal ulcer, post-endoscopic therapy bleeding, and others were analyzed as variables.

Patients underwent pre-colonoscopy bowel preparation using 2–4 L of a solution containing polyethylene glycol or enema on the day of the colonoscopy, considering the patient’s condition. The endoscopic cap and water jet device were used based on the practitioner’s preference and decision. All colonoscopies were performed using an electronic video endoscope (Olympus Optical, Tokyo, Japan, or Fujifilm Corporation, Tokyo, Japan).

The final diagnosis was obtained by inpatient examinations after hospitalization due to the suspicion of acute lower gastrointestinal bleeding. SRH type was categorized as active or non-active bleeding (adherent clot and/or visible vessel). Endoscopic hemostasis included clipping, band ligation, detachable snare ligation, electrocoagulation, and hypertonic saline-epinephrine, which were selected at the endoscopist's discretion and in accordance with hospital policy.

Statistical analysis

Pearson's chi-square, Fisher's exact, and Mann–Whitney U tests were used to analyze surgery-associated factors. In addition, odds ratios and 95% confidence intervals (CI) were calculated by uni- and multivariate logistic regression analysis and for risk factors associated with surgery needs. Variables with a univariate significance of P less than 0.05 were entered into the multivariate analysis. The data were analyzed using SPSS statistical software version 25 (IBM, New York, NY, USA), and a P value < 0.05 was considered significant. In the multivariate analysis of surgery-related risk factors, explanatory variables that were significantly different P < 0.05 in the univariate analysis were used so that the number of explanatory variables was statistically valid according to the number of events.

Results

Characteristics and outcomes of all surgery cases

Surgery was performed in 142 cases during hospitalization, and data analysis was performed in 136 cases, excluding six with insufficient clinical information. Of the 49 hospitals participating in the study, 29 (59%) had enrolled patients who were undergoing surgery during their hospital stay. The rate of surgery for ALGIB at each of the 29 hospitals ranged from 0.25 to 5.6%, with a median of 1.16%. The indications for surgery were CDB in 33 cases (24.3%), colorectal cancer in 30 (22.1%), small bowel bleeding in 22 (16.2%), adverse events of endoscopy and interventional radiology performed for bleeding in ten (7.4%), hemorrhoid bleeding in seven (5.1%), inflammatory bowel disease in six (4.4%), intestinal perforation with bleeding in five (3.7%), intestinal ischemia in five (3.7%), rectal ulcer bleeding in three (2.2%), and the others in 15 (11.0%) (Fig. 2). Regarding small bowel bleeding, surgery rates were significantly lower (p = 0.03) at 2% (1/54) in patients with endoscopic hemostasis as compared with 12% (21/179) in those without endoscopic hemostasis.

Fig. 2
figure 2

Causative disorders for surgery (n = 136). *Meckel's diverticulum (n = 9), vascular lesion (n = 5), cancer (n = 2), non-specific ulcerative lesion (n = 2), Behçet’s disease (n = 1), malignant lymphoma (n = 1), intestinal invasion from hepatoma (n = 1), and submucosal lesion (n = 1). **colonic perforation (n = 4) and small bowel perforation (n = 1). ***gangrenous ischemic colitis (n = 2), small bowel ischemia (n = 1), non-occlusive mesenteric ischemia (n = 1), and stricture type of ischemic colitis (n = 1). ****incidentally diagnosed lesion not related to hematochezia (n = 4), upper gastrointestinal bleeding (n = 2), ovarian cancer with colonic invasion (n = 2), rectal bleeding of unknown origin (n = 2), cytomegalovirus enteritis (n = 1), arteriovenous malformation (n = 1), bowel intussusception (n = 1), rectal injury (n = 1) and perirectal abscess (n = 1)

Full size image

In the classification according to purpose, surgeries for hemostasis to severe refractory bleeding were performed in 87 cases (64%), and surgeries mainly not for hemostasis were performed in 49 (36%) (Table 1). Emergency surgeries for hemostasis were performed in 19 patients (22%), including those for CDB (n = 9, 47%), small bowel bleeding (n = 5, 26%), and rectal hemorrhagic ulcer (n = 2, 11%). Elective surgeries for hemostasis were performed in 68 patients (78%), including those for CDB (n = 24, 35%), small bowel bleeding (n = 14, 21%), and colorectal cancer (n = 11, 16%). Emergency surgeries mainly not for hemostasis included those for refractory ulcerative colitis, necrotic ischemic enteritis, and intussusception with bleeding. Elective surgeries, mainly not for hemostasis, included those for colorectal cancers that had stopped bleeding but required surgical resection. Surgical procedures included total or subtotal colectomy in eight cases (6%), segmental colectomy in 78 cases (57%), segmental resection of small bowel in 21 cases (15%), and non-resective surgeries in 29 cases (21%) (Table 1). Intraoperative adverse events were observed in only one case (0.7%) (transfusion-induced anaphylaxis occurred during surgery), postoperative adverse events in 28 cases (20%), surgical adverse events in 13 cases (9.6%), and general adverse events in 11 cases (8.1%).

Table 1 Characteristics of surgery
Full size table

Surgeries for hemostasis were performed in 87 cases, and confidence levels for preoperative identification of bleeding site were definitive in 50 patients (57%), presumptive in 33 (38%), and obscure in four (5%). Postoperative rebleeding was observed in 14 of 87 cases (16%), and the median length until postoperative rebleeding was 99 days (Inter quartile range: 32–652). The causes of rebleeding in the six cases with definitive preoperative identification of the bleeding site were mostly bleeding from the anastomotic site (3/6) or anal rebleeding from a hemorrhoid or perianal abscess (2/6). Only one case involved rebleeding from the colon diverticula after right hemicolectomy in cases with definitive preoperative identification of the bleeding site of the ascending colon cancer. The causes of rebleeding in the eight cases with presumptive or obscure preoperative identification of the bleeding site were CDB in three, small bowel bleeding in two, anastomotic bleeding in one, and obscure bleeding in two (Table 2).

Table 2 Outcomes of surgery
Full size table

Thirty-day mortality rate was 1.5% (2/136) in patients who underwent surgery, and mortality during hospitalization in all surgery cases was 2.9% (4/136) (Table 2).

Assessment of associated risk factors of surgery using multivariate regression (surgery vs. non-surgery)

In a multivariate analysis, risk factors significantly associated with surgery were transfusion of six or more units (OR 2.50; 95% CI 1.58–3.95, P < 0.001), in-hospital rebleeding (OR 2.67; 95% CI 1.66–4.29, p < 0.001), small bowel bleeding (OR 7.87; 95% CI 4.07–15.2, p < 0.001), colorectal cancer (OR 23.2; 95% CI 12.7–42.3, p < 0.001), and hemorrhoids (OR 5.30; 95% CI 2.17–13.0, p < 0.001). On the contrary, endoscopic hemostasis (OR 0.33; 95% CI 0.16–0.68, p = 0.003) was the factor negatively related to surgery (Table 3).

Table 3 Risk factors associated with surgery
Full size table

Discussion

This study was the first to identify risk factors associated with surgery for ALGIB in a large nationwide cohort. Transfusion need ≥ 6 units, in-hospital rebleeding, small bowel bleeding, colorectal cancer, and hemorrhoids conferred significantly higher odds of surgery, whereas endoscopic hemostasis had significantly lower odds in multivariate analysis. Transfusion need ≥ 6 units and in-hospital rebleeding after primary inpatient care, including hemostatic intervention, indicate severe refractory bleeding and are understandable as a surgery-related risk. Colon cancer and hemorrhoids with overt hematochezia upon admission require consequential surgery owing to the nature of the disease. On the other hand, it is clinically noteworthy that small bowel bleeding was a significant risk factor for surgery (OR 7.87, 95% CI 4.07–15.2, p < 0.001). Small intestinal bleeding presents as either obscure gastrointestinal bleeding or overt bleeding with hematochezia. The latter overt small bowel bleeding was the third top cause (16.2%) for surgery in this cohort of acute hematochezia. Similarly, a recent report enrolling 87 surgery cases of 1198 ALGIB patients [10] documented that small bowel disease was the second most common cause of surgery. When small bowel bleeding was considered, the surgery rate was significantly lower (p = 0.03) at 2% (1/54) in patients with endoscopic hemostasis as compared with 12% (21/179) in those without endoscopic hemostasis. The test to diagnose small bowel bleeding is not available in all hospitals. Taken together, however, these results suggest that small bowel exploration is essential in acute hematochezia when the bleeding site cannot be unlocalized with standard examinations, because small bowel bleeding is a major cause of ALGIB and has a high risk of requiring surgery.

In our cohort, colonoscopy was performed in 88% with a high diagnostic yield (94.9%) [20, 21]. In 27% of colonoscopy examinations, endoscopic hemostasis was performed and conferred significantly low odds of surgery (OR, 0.32; 95% CI 0.16–0.68, p = 0.003). Taken together, colonoscopy to identify the bleeding source and subsequent endoscopic hemostasis have contributed to better management of ALGIB by reducing surgical intervention. Consequently, surgery was performed in only 1.2% of 10,342 acute hematochezia cases, in which colonoscopy and endoscopic hemostasis were done at high rate of 88% and 27%, respectively. Our literature review showed that surgery rates ranged from 34 to 46% between 1980 and 1999 [12,13,14,15], from 14 to 19% in the 2000s [16,17,18] and from 0.2 to 18% with a median of 4.8% after 2010 [6,7,8,9,10,11]. Surgery for ALGIB has been declining over time, probably due to the improvement of total medical management, including diagnosis of a bleeding source and non-surgical hemostasis. Grayer C. et al. [18] analyzed 1,112 ALGIB cases and demonstrated that endoscopic hemostatic control significantly increased from 1% in 1988–1997 to 4.4% in 1998–2006, with a corresponding decrease in surgical control from 22.6 to 16.6% in the same time periods.

Detailed data on postoperative rebleeding after surgery for ALGIB have not been reported to date, and this nationwide study was the first to focus on this issue. Postoperative rebleeding rates were 16% (14/87) in total, 22% (8/37) in patients with presumptive or obscure preoperative identification of the bleeding source, and 12% (6/50) in those with definitive identification. In 6 postoperative rebleeding cases with definitive preoperative identification, three were anastomosis bleeding and one was bleeding from primary anal lesions, meaning that only 2 of 50 cases (4%) were postoperative bleeding from sources unknown preoperatively. On the other hand, in 8 postoperative rebleeding cases with presumptive or obscure preoperative identification, seven were postoperative bleeding from sources unknown preoperatively. These results demonstrate that definitive preoperative identification of a bleeding source is pivotal for reducing postoperative rebleeding and again support the British guideline that laparotomy without localization of the source of the bleeding should be avoided.

In this cohort, total or subtotal colectomy was 6% of 136 surgical procedures, and emergency surgery for hemostasis was 14% of those, indicating that localized colectomy or segmental intestinal resection has been electively performed in most cases. A 30-day mortality rate was 1.5% in all surgery cases of the above operative profile, comparable to the rate of 0.8% in patients without surgery. In contrast, a 30-day mortality rate was higher at 12.2% in a recent cohort of 1614 emergency surgery cases; of those, total or subtotal colectomy was 24% [11]. As shown in the British guideline [25], emergency total or subtotal colectomy for ALGIB has a high 30-day mortality even in recent years and should be avoided by extensive exploration of bleeding source and non-surgical hemostasis.

The present study has shown that the fourth cause for surgery was complications of diagnostic and therapeutic intervention (7%). The rate of TAE-related complications requiring surgery in this study was 2.1% (3/143), while the rates were previously reported from 3.6 to 13% [26,27,28], showing that surgery could be indicated due to TAE-related complications at a certain unignorable percentage. The rate of colonoscopy-related perforation requiring surgery in this study was 0.054% (6/11056, 5.4 per 10,000 colonoscopies) in total, 0.040% (3/7486) in diagnostic colonoscopy, and 0.084% (3/3570) in endoscopic hemostasis. The recent systematic review [29] demonstrated that the pooled rate of perforations among 10,328,360 colonoscopies was 5.8 per 10,000 colonoscopies (95% CI, 5.7–6.0). The perforation rate in this study was below the lower value of 95% CI shown in the systemic review, suggesting that colonoscopy could be performed in ALGIB without the increase in perforation risk. The perforation rate of endoscopic hemostasis in this study was 0.084%, considerably lower than the rate of TAE-related perforation (2.1%). Given together with the benefit in the significant reduction of surgery needs and the relatively low risk of perforation, endoscopic hemostasis can be selected as the first-line hemostasis.

The present study had some limitations. First, because of the retrospective design of this study, there could have been missing data and the possibility of bias. However, no missing data were available concerning diagnosis, procedures, interventions, or outcomes, and items with missing values and their rates in our cohort were comparably low, as shown previously [20, 21]. After efforts were made to minimize missing data, multivariate analysis was performed using complete-case analysis. Second, this was a large-scale cohort collected from 49 hospitals, but there may be some bias regarding region, hospital size, and the nature of the institution. Third, we cannot present the exact number of small bowel bleeding cases diagnosed with balloon-assisted endoscopy or capsule endoscopy because we did not collect this information. Fourth, the multivariate analysis on surgery-related risk factors, the possible presence of potential confounders other than those statistically analyzed, and the setting of cut-off values for the risk factors may alter the results.

Conclusions

Our large-scale cohort study has comprehensively elucidated the characteristics, outcomes, and risk factors of surgery for acute hematochezia. Extensive exploration including the small bowel to identify the source of bleeding and endoscopic hemostasis may reduce unnecessary or high-risk surgery and improve the management of ALGIB.