Introduction

Colorectal cancer (CRC) is one of the most common malignancies and a major cause of cancer-related death worldwide [1]. Although the incidence of CRC in developed countries has declined, it is increasing rapidly in developing countries [2]. Malignant colorectal obstruction (MCO) caused by advanced CRC and other metastatic cancers [3] often requires emergency surgical intervention such as colostomy, colorectal resection, or Hartmann’s procedure [4, 5]. These interventions are associated with a high risk of complications, some of which are life threatening. According to reports, these surgical interventions were associated with mortality and morbidity rates of 15–20 and 40–50 %, respectively [4, 6]. Many patients with MCO have multiple metastases or an unresectable primary malignancy. The ostomy that relieves MCO will never be reversed for those patients and the health-related quality of life will be compromised dramatically for their expected life span [7].

The metallic self-expanding stent (SEMS) introduced by Dohmoto et al. [8] in 1991 has been used for the palliative treatment of MCO [3], gastroduodenal obstruction [9], and bile duct obstruction [10]. In past decades, SEMS was used for treating MCO as a bridge to selective surgery for patients with curable or resectable CRC [11–13], or as an alternative for ostomy or Hartmann’s procedure for MCO in patients with advanced colorectal malignancy [3]. Although SEMS can treat MCO effectively and avoid surgical intervention in most patients with advanced CRC, the short-term (within 30 days after surgery) and long-term (from 30 days after surgery to last follow-up) mortality and morbidity rates are major concerns. In theory, SEMS requires minimal hospitalization and can improve quality of life, but it may reduce survival when the original tumor cannot be removed. Most importantly, SEMS is more tolerable than surgery and can be applied to the patients with cardiovascular or renal dysfunction; however, stent-related perforation can cause life-threat complication and even death. A randomized controlled trial by van Hooft et al. [14] indicated a high incidence of perforation of up to 18 % in the short term and 36 % in the long term. Stent migration and re-obstruction also require a repeat procedure. The oncologic effects of a stent vs. surgery remain uncertain, especially for patients with a resectable metastatic malignancy.

Using a system review and meta-analysis, we compared the outcomes of SEMS alone vs. surgical intervention for patients with an MCO caused by advanced colorectal cancer, in terms of short-term and long-term complications, mortality, and time of hospitalization.

Materials and methods

This study was conducted according to the PRISMA guidelines [15] for reporting system review and meta-analysis.

Literature search

We retrieved the following databases: MEDLINE (1990–31 August 2011), EMBASE (1990–31 August 2011), Cochrane controlled trials registry (updated to 31 August 2011) and Wanfang (Chinese local database, 1990–31 August 2011). The following key words were jointly used: obstruction, stent OR stenting, colorect* OR colon* OR colon OR rectum, randomized OR (randomized trial) OR (randomized controlled trial), (controlled trial) OR (clinical trial). No language limitations were applied. We also manually searched the reference lists of all eligible studies and reviews closely related to our topic, published in the last 5 years.

Inclusion and exclusion criteria

The following criteria were used for the selection of eligible studies:

  1. 1.

    Study design: controlled trials with humans, including randomized controlled and non-randomized controlled trials.

  2. 2.

    Study population: patients with MCO caused by advanced malignancy.

  3. 3.

    Intervention: SEMS alone vs. any surgical treatments.

  4. 4.

    Outcomes: short-term complications, long-term complications, mortality and time of hospitalization. (Only studies including two or more of these outcome measures were included).

The study was excluded, if the outcomes reported did not include the two arms; if they were impossible to calculate indirectly; or if the SEMS was used as a bridge to selective surgery.

Study selection and data extraction

Two authors (T.W. Liang and Y. Sun) identified possible eligible studies independently and then decided on their true eligibility after discussion together. The detailed information leading to study selection or exclusion is presented in the Fig. 1. Briefly, after exclusion of duplications cited among the databases, there were 132 publications that potentially fitted our topic. Through screening the titles and abstracts, 104 studies were excluded. The full texts of 28 studies were then reviewed and 19 were excluded for one or more of the following reasons: the study was not a clinical controlled trial; SEMS was used as a bridge to selective surgery; or the study was irrelevant to our topic. Finally, nine primary studies were analyzed [3, 14, 16–22].

Fig. 1
figure 1

Flow chart detailing information on the primary study selection and reasons for exclusion

Full size image

The following details were extracted from the original articles: the first author’s name; the publication year; the simple sizes in two arms; the study design; the obstructive parameters; the outcome parameters, including perforation and death; and information for quality assessment.

Quality assessment

The quality of primary studies, including randomized and non-randomized controlled studies, was assessed by the Newcastle–Ottawa Scale (http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp) with some modifications to fit our study [13, 23]. The modified scale consisted of four parts: patient selection, comparability of study groups, assessment of outcome, and study design (Supplement 1). A maximum of three, four, two and two stars were awarded to above four items, respectively. If the study achieved six or more stars, it was regarded as being of high quality.

Definition of outcomes

The primary outcomes in this meta-analysis were mortality within 30 postoperative days and postoperative complications. The major complications related to SEMS were colorectal perforation, stent migration and re-obstruction. Short-term complications were defined as those that manifested within 30 postoperative days, and long-term complications were defined as those that manifested from 30 postoperative days to the final follow-up or death. The secondary outcomes in this study were the success rate of stent replacement for relieving the MCO and the length of hospitalization.

A lack of data in primary studies or great heterogeneity between patients among original studies precluded us from studying other outcomes such as intervention time, readmission, or survival time.

Statistical analysis

The odds ratio (OR) and 95 % confidence interval (95 %CI) were used to judge the strength of the association for dichotomous variables, and the weighted mean difference (WMD) was used for continuous variables. The Z test was also used to determine significance (P < 0.05). Because of inherent heterogeneity among primary studies originating from different inclusion criteria and variable risk profiles for patients, a random-effect model [24], which could provide a more conservative estimate, was used to combine the data. If continuous data were reported as the median with a range, the mean with standard deviation (SD) was estimated by the method introduced by Hozo et al. [25]. The following situations were impossible for combined analysis: continuous data not presented as means with standard deviation or median with range; in dichotomous data, zero cells reported as outcomes in both the SEMS and surgical groups.

To keep clinical homogeneity and prove the reliability of our study, subgroup analyses were performed by the following variables: (1) study design (prospective or retrospective); (2) types of surgical treatments. The type-I comparison was defined as SEMS vs. colostomy without tumor resection and the type-I/II comparison was defined as SEMS vs. colostomy without tumor resection or surgical treatment with tumor resection. This was because either colostomy or other surgical treatments with tumor resection were carried out for patients in the surgical arm of five original articles [14, 18, 19, 21, 22]; (3) stage IV CRC; (4) left-sided CRC; and (5) high-quality studies.

Publication bias was tested by several methods. Asymmetry of a funnel plot indicated possible publication bias. Begger and Egger’s linear regression tests [26, 27] were used to determine potential publication bias, and P < 0.05 indicated significance. All statistical tests were done by STATA software (version 10.0; Stata Corporation, College Station, TX) with two-sided P values.

Results

Primary characteristics of the included studies

Tables 1 and 2 summarize the basic information in the collective nine primary studies [3, 14, 16–22]. Our study included 410 patients. The sample sizes ranged from 21 to 123. The participants were mainly from western countries, but with one study of Chinese patients [19] and one of Japanese patients [3]. SEMS and surgery were performed to relieve MCO in 195 (47.6 %) patients and 215 patients (52.4 %), respectively. MCO was caused exclusively by advanced colorectal malignancy, especially by stage-IV CRC. The patients in six studies had left-sided MCO only [14, 16, 17, 19, 20, 22]. The age and gender distributions between the SEMS and surgical groups were comparable [WMD (95 %): 2.79 (−1.73, 7.31), P = 0.23 for age and OR (95 %): 0.93 (0.61, 1.42), P = 0.75 for gender]. Three studies [14, 16, 17] were randomized controlled trials and six [3, 18–22] were clinical controlled trials. Seven studies [3, 14, 16–19, 22] were awarded more than six stars and regarded as high quality in the subgroup analysis.

Table 1 Primary characteristics of the nine studies included in our analysis
Full size table
Table 2 Outcomes and quality assessment reported in the nine primary studies
Full size table

Treatment success

MCO was effectively relieved by surgery in all (100 %) cases and by SEMS in 172 (93.9 %) of 183 cases among 195 patients. The main reasons for failure of SEMS were as follows: the stent was difficult to place because of tortuous flexure of the colorectum, especially when the flexure was splenic or rectosigmoid [17, 20]; the site of obstruction was low [20]; or the colon anatomy had been altered by the tumor and a guidewire could not be passed through the obstruction [18].

Major complications of SEMS

The major complications of SEMS were colorectal perforation, stent migration, and re-obstruction. Five studies [3, 14, 18, 19, 22] reported five cases of perforation in the short term, and four studies [3, 14, 18, 22] reported eight cases of perforation in the long term. The incidence of perforation in these primary studies varied greatly. van Hooft et al. [14] reported the highest short- and long-term incidences of 18 and 36 %, respectively, while other studies reported incidences of 0–8 %. The overall incidences of short- and long-term perforation were 3.7 % (5 cases among 135 patients) and 7.6 % (8 cases among 106 patients), respectively. Although stent migration did not always cause serious adverse events, it required replacement. The incidence of stent migration was reported to range from 5.5 to 10 %, with an overall incidence of 8.9 % (12 cases/135 total patients). Interestingly, none of the primary studies mentioned the incidence of re-obstruction. Table 3 gives details about the complications of SEMS. The three studies [14, 19, 22] that focused on left-sided MCO reported incidences of short- and long-term perforation as well as stent migration of 2.4, 12.3, and 6.3 %, respectively.

Table 3 Main complications of stent placement reported in the nine primary studies
Full size table

Quantitative analysis of complications

All primary studies provided information for the analysis of short-term complications. The overall incidences were 26.2 % (51 cases among 195 patients) after SEMS and 34.5 % (74 cases among 215 patients) after surgery. The results of combined analysis showed no significant difference between the two groups [OR (95 %): 0.83 (0.39, 1.79), P = 0.22; Fig. 2a)]. Seven studies [3, 14, 17, 18, 20–22] reported on the long-term complications of SEMS and surgery, and one [17] did not. Therefore, six primary studies [3, 14, 18, 20–22] provided information for combined analysis. The overall incidences were 16.1 % (25 cases among 155 total patients) after SEMS and 8.1 % (14 cases among 173 patients) after surgery. The patients who underwent SEMS had a significantly higher risk of long-term complications than those who underwent surgery [OR (95 %CI): 2.34 (1.07, 5.14), P = 0.03; Fig. 2b)]. Seven primary studies [3, 14, 17, 18, 20–22] reported on the incidence of complications overall. The incidences in the two groups were 43.9 % (68 cases among 155 total patients) and 45.1 % (78 cases among 173 total patients), respectively. The combined results indicated that patients treated with SEMS and those treated with surgery shared a similar risk of overall complications [OR (95 %CI): 1.27 (0.58, 2.77), P = 0.56].

Fig. 2
figure 2

Forest plot showing combined analyses of the short-term (a) and long-term (b) complications using a random-effect model. The results indicated that SEMS was associated with an increased risk of long-term complications rather than short-term complications compared with surgical intervention

Full size image

Six [14, 16, 17, 19, 20, 22], three [14, 20, 22] and four [14, 17, 20, 22] studies provided information on the quantitative analyses of short-term, long-term, and total complications of left-sided MCO, respectively, but none reached significance [OR (95 % CI): 1.10 (0.41, 2.92), P = 0.85 for short-term complication; 3.29 (0.81, 13.38), P = 0.10 for long-term complication; 2.04 (0.61, 6.79), P = 0.25 for total complication].

Quantitative analysis of length of hospitalization

All primary studies reported on the overall length of hospitalization, and seven studies [3, 14, 16, 18–20, 22] were available for combined analysis. The patients who underwent SEMS required significantly shorter hospitalization than those who underwent surgery [WMD (95 % CI): −6.07 (−8.40, −3.74), P < 0.01; Fig. 3). Five studies [14, 16, 19, 20, 22] provided information for the analysis of the length of hospitalization required for the treatment of left-sided MCO. Patients treated with SEMS had significantly shorter hospitalization than those treated with surgery [WMD (95 %CI): −4.94 (6.90, −2.97), P < 0.01].

Fig. 3
figure 3

Forest plot showing combined analyses of time of hospitalization using a random-effect model. The results indicated that SEMS was significantly associated with shorter time of hospitalization than surgical intervention

Full size image

Quantitative analysis of mortality

The overall mortality rates in the seven studies [3, 14, 18–22] were 7.1 % (12 deaths among 169 patients) for SEMS and 11.6 % (22 deaths among 189 patients) for surgery. These combined results indicated that both procedures had similar risks of mortality [OR (95 % CI): 0.60 (0.27, 1.34), P = 0.22; Fig. 4)]. Similarly, based on the four studies [14, 19, 20, 22] there was no significant difference in the mortality of patients with left-side MCO according to whether they were treated with SEMS or surgery [OR (95 %CI): 0.66 (0.26, 1.70), P = 0.39].

Fig. 4
figure 4

Forest plot showing combined analyses of mortality using a random-effect model. The results indicated that there was no statistical significance between the SEMS and surgery groups

Full size image

Other subgroup analyses

Table 4 shows the results of comprehensive subgroup analyses. Overall complications and mortality in the two groups did not differ in any of the subgroup comparisons; however, the length of hospitalization was significantly shorter for SEMS than for surgery in all subgroup analyses except for SEMS vs. colostomy [WMD (95 %CI): −7.62 (−22.38, 7.17), P = 0.31].

Table 4 Subgroup analyses of postoperative complications, mortality, and hospitalization
Full size table

Publication bias test

We inspected the symmetry of funnel plots and confirmed no publication bias for comparisons of short-term, long-term and overall complications, hospitalization and mortality (Fig. 5a–e). The results of Begger and Egger tests also indicated no publication bias, which might be attributed to the low power of statistics [26, 27] (P Begger = 0.92 and P Egger = 0.94 for short-term complication; P Begger = 0.26 and P Egger = 0.88 for long-term complication; P Begger = 1.00 and P Egger = 0.69 for overall complications; P Begger = 0.55 and P Egger = 0.67 for time of hospitalization; P Begger = 0.92 and P Egger = 0.94 for hospitalization; P Begger = 1.00 and P Egger = 0.84 for mortality).

Fig. 5
figure 5

Funnel plots for publication bias: a for short-term comparison; b for long-term complications; c for overall complications; d for time of hospitalization; and e for mortality

Full size image

Discussion

Advanced unresectable CRC is the major cause of MCO, although it has also been associated with other malignancies such as metastatic gastric cancer and metastatic ovarian cancer [3]. It has been reported that about one-third of patients with CRC suffer the symptoms of MCO [28] at some stage, and the most common side of obstruction is the left [29] because most CRC arises in the left side, where the lumen narrows from the ascending colon into the rectum. Traditionally, two-step surgery; namely, Hartmann’s procedure and reversal of ostomy, is needed for resectable CRC, and ostomy is needed to treat MCO in patients with advanced unresectable CRC. These invasive procedures are associated with huge morbidity and mortality risks. Since Dohmoto et al. [8] introduced the metallic stent in 1991, SEMS has been used widely to relieve gastrointestinal obstruction. In 2007, Watt et al. [30] conducted a system review to compare SEMS with surgery for relieving MCO in patients with resectable or unresectable colorectal malignancy. Based on published studies, SEMS is concluded to be safe and effective to overcome left-sided MCO, with a shorter hospital stay and a lower rate of serious events than surgery. Recently, a meta-analysis by Zhang et al. [13] confirmed that SEMS as a bridge to selective surgery increased the chance of primary anastomosis and decreased post-procedure complications and the need for stoma. SEMS has also been widely used to relieve MCO in patients with primary unresectable malignancy or multi-metastatic malignancy. Similarly, the system review indicated that SEMS is effective for relieving MCO in patients with incurable colorectal malignancy.

This meta-analysis indicated that performing SEMS instead of surgery for MCO caused by advanced unresectable malignancy was associated with shorter hospitalization and quicker recovery. Moreover, SEMS and surgical interventions such as colostomy and Hartmann’s procedures had similar risks of short-term complications and mortality; however, SEMS had a higher risk of long-term complications. Although most of these patients with a stoma would never undergo reversal of the stoma considering their survival prediction [31], the complications of stoma reversal should nevertheless be considered. All of the primary studies reported only complications of colostomy or Hartmann’s procedure. It is possible that the risks of morbidity and mortality are much higher for two-step surgery, such as Harmann’s procedure followed later by reversal of the stoma, than for SEMS [32–34]. Furthermore, the subgroup analysis for left-side MCO revealed no significant differences in the long-term and overall complications between the SEMS and surgery groups.

One of the greatest advantages of SEMS for the palliative treatment of MCO is that it eliminates the need for a stoma, which requires a high level of nursing care and compromises both psychology- and physics-related quality of life. The most common complication of stoma is dermatitis [35], while serious complications include parastomal hernia, sclerosis, and stenosis [32, 35]. SEMS is less invasive and consequently better tolerated than surgery, with less suffering. Thus, patients with organ dysfunction, such as renal or heart failure, for whom the risks of surgical morbidity and mortality were too high, were good candidates for SEMS. Cost-effectiveness was also an important advantage of SEMS [21].

Some issues of SEMS should be noted. First, colorectal perforation is a life-threatening complication, which requires emergency surgery. van Hooft et al. [14] reported the highest rates of colorectal perforation: 18 % for short-term perforation and 36 % for long-term perforation, respectively. Another four studies reported lower incidences of perforation at about 0–7.6 %. In the present study, the overall incidences of perforation were 3.7 % in the short term and 7.6 % in the long term. The reasons for perforation are unclear, but the following may be contributing factors: First, persistent high pressure on the colon wall. Insufficient dilation of an obstructive tumor can cause persistent high pressure on the colorectal wall at the site of the proximal tumor, which could lead to ischemia and necrosis. A large stent diameter could also compress the colon wall, predisposing to perforation. Second, optimal design and material of the stent might contribute to a low risk of perforation. van Hooft et al. [14] stated that an association between the incidence of perforation and the design of a new stent could not be excluded. Theoretically, less inflammatory reaction caused by better stent material might help prevent perforation. Third, prolonged and severe obstruction is associated with colonic edema and dysfunction of microvessel blood circle and this pathology would predispose to colonic perforation. Finally, the anatomic location of the stent; for example, if it is in the curved sigmoid colon, might account for a high risk of perforation. This notion was supported by a large review article by Sebastian et al. [36].

Our meta-analysis revealed a higher risk of long-term complications for SEMS than for surgery. Although only a small proportion of the long-term complications of SEMS was life-threatening, we should nevertheless take this into consideration. Moreover, although the incidence of long-term complications after surgery appeared to be low, some serious complications of surgery, such as wound hernia and adhesive intestinal obstruction, always occurred more than 2 years later, whereas the life expectancy of patients with MCO caused by advanced malignancy was often <2 years. Besides, as all of the primary studies focused on SEMS rather than on surgical treatment, it is likely that some of the complications of surgery were omitted. For example, dermatitis is a common complication of colostomy, but this was not mentioned in any primary studies.

Our meta-analysis did not include the oncologic outcomes of SEMS as the great heterogeneity of the patients was reported in the primary studies. The chemoradiotherapy strategies were also different. In theory, SEMS should be the optimal palliative treatment for MCO in patients with advanced unresectable malignancy. Three primary studies reported survival times after colostomy without tumor resection vs. SEMS. Xinopoulos et al. and Johnson et al. [17, 20] reported slightly longer survival after SEMS than colostomy, but without significance. On the other hand, Tomiki et al. [3] reported longer survival after colostomy than after SEMS, but again the difference was not significant. Despite limited data on how SEMS affects survival, it was evident that SEMS had a similar influence to colostomy on the overall survival time of patients with MCO caused by advanced unresectable malignancy. Thus, the effect of SEMS on the prognosis of these patients should be assessed by larger randomized controlled trials.

Like other meta-analyses, this study had potential bias and limitations [10, 13]. First, as the number of patients was small, with relatively low statistical power, the results should be interpreted with caution. Even so, this study provides useful information about treating MCO with SEMS. Second, the literature search strategy was designed before the initiation of this study, and potentially eligible studies were identified by computer-based and manual search; however, it is possible that a few eligible studies were not included. Third, the results of this meta-analysis, a type of retrospective study, could have been distorted by possible selection bias [37]. Fourth, we did not investigate the oncologic effect of SEMS vs. surgical intervention, as great heterogeneity for patients and oncologic treatments were reported in primary studies.

In summary, our meta-analysis showed SEMS to be as effective as surgery for MCO in patients with advanced unresectable colorectal malignancy, with shorter hospitalization, quicker recovery, and comparable risk of short-term complications and mortality. Considering the limitations of this study, larger sample-sized randomized controlled trials are needed to further confirm our findings.