Covered versus Uncovered Self-Expandable Metal Stents for Managing Malignant Distal Biliary Obstruction: A Meta-Analysis

Aim To compare the efficacy of using covered self-expandable metal stents (CSEMSs) and uncovered self-expandable metal stents (UCSEMSs) to treat objective jaundice caused by an unresectable malignant tumor. Methods We performed a comprehensive electronic search from 1980 to May 2015. All randomized controlled trials comparing the use of CSEMSs and UCSEMSs to treat malignant distal biliary obstruction were included. Results The analysis included 1417 patients enrolled in 14 trials. We did not detect significant differences between the UCSEMS group and the CSEMS group in terms of cumulative stent patency (hazard ratio (HR) 0.93, 95% confidence interval (CI) 0.19–4.53; p = 0.93, I2 = 0%), patient survival (HR 0.77, 95% CI 0.05–10.87; p = 0.85, I2 = 0%), overall stent dysfunction (relative ratio (RR) 0.85, M-H, random, 95% CI 0.57–1.25; p = 0.83, I2 = 63%), the overall complication rate (RR 1.26, M-H, fixed, 95% CI 0.94–1.68; p = 0.12, I2 = 0%) or the change in serum bilirubin (weighted mean difference (WMD) -0.13, IV fixed, 95% CI 0.56–0.3; p = 0.55, I2 = 0%). However, we did detect a significant difference in the main causes of stent dysfunction between the two groups. In particular, the CSEMS group exhibited a lower rate of tumor ingrowth (RR 0.25, M-H, random, 95% CI 0.12–0.52; p = 0.002, I2 = 40%) but a higher rate of tumor overgrowth (RR 1.76, M-H, fixed, 95% CI 1.03–3.02; p = 0.04, I2 = 0%). Patients with CSEMSs also exhibited a higher migration rate (RR 9.33, M-H, fixed, 95% CI 2.54–34.24; p = 0.008, I2 = 0%) and a higher rate of sludge formation (RR 2.47, M-H, fixed, 95% CI 1.36–4.50; p = 0.003, I2 = 0%). Conclusions Our meta-analysis indicates that there is no significant difference in primary stent patency and stent dysfunction between CSEMSs and UCSEMSs during the period from primary stent insertion to primary stent dysfunction or patient death. However, when taking further management for occluded stents into consideration, CSEMSs is a better choice for patients with malignant biliary obstruction due to their removability.


Introduction
Most patients with a malignant biliary obstruction cannot undergo operations because of the advanced stage of the obstruction. Transhepatic or endoscopic stent insertion is the gold standard for palliative treatment to relieve biliary obstructions and improve quality of life [1][2][3]. Studies have shown that both plastic stent insertion and self-expandable metal stent (SEMS) insertion effectively relieve jaundice and improve quality of life [4,5]. However, stent occlusion is the primary clinical problem associated with endoprostheses, leading to re-intervention [6]. Typically, plastic stents become occluded with sludge because of their small diameter. Enlarging the diameter seems to be a solution, but the diameter is limited by the endoscope delivery system [1,7]. Uncovered SEMSs (UCSEMSs), which are only 3.5-4 mm in diameter, can expand to 8-10 mm in the bile duct after being released from the delivery system. Compared with plastic stents, UCSEMSs exhibit longer patency durations and lower occlusion rates, reduce the length of hospital stay and decrease the need for periodic stent exchange [8][9][10]. However, tumor ingrowth via the metal mesh has been observed in patients with UCSEMSs, which is one of the main causes of UCSEMS occlusion [11]. To prevent tumor ingrowth and avoid re-intervention, SEMSs covered with a non-porous membrane were developed.
Several trials have compared UCSEMSs and covered SEMSs (CSEMSs), but the conclusions have been inconsistent [12][13][14][15][16][17][18][19]. Moreover, two recent meta-analyses produced different conclusions [20,21]. Thus, we performed a meta-analysis to reassess the efficacy of UCSEMSs compared with CSEMSs. This study specifically compares the cumulative stent patency rate, patient survival, stent dysfunction, overall complication rate and biliary drainage efficacy after CSEMS or UCSEMS placement to identify the more favorable option for palliative treatment of a distal malignant biliary obstruction.

Eligibility criteria
Types of studies: Randomized controlled trials (RCTs) comparing the efficacy of UCSEMSs and CSEMSs in treating distal malignant biliary obstruction were included. Non-and quasirandomized trials were excluded.
Types of interventions: Both endoscopic and percutaneous approaches for stent placement for managing distal malignant biliary obstruction were considered. Both types of stents had to be commercially available and not handcrafted.
Types of participants: The study population was as follows: >18 years old, without gender restrictions, and diagnosed with obstructive jaundice caused by an unresectable tumor (not a hilar tumor).
All of the trials had been published in English.

Search methods used to identify the studies
We conducted a comprehensive search of electronic databases (the EMBASE, PubMed, Science Citation Index Expanded and Cochrane Library databases) for trials from 1980 to May 2015.
The search strategy was based on MeSH terms combined with key words. The detailed search strategies are described in S1 Table. Outcome choice Cumulative stent patency and patient survival were considered as the primary outcomes. In this paper, only primary stent patency was analyzed. Primary stent patency is associated with an absence of recurrent symptomatic biliary obstruction and is equivalent to the duration from primary stent placement to stent dysfunction. Overall stent dysfunction, the overall complication rate and a change in serum bilirubin were secondary outcomes.

Data collection
Two authors blinded to the authors and institutions of the search results selected the studies by following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) process. Disagreements were solved through discussion with a third author. The reasons for exclusion were recorded. Two authors independently extracted the data from the included studies, and disagreements were discussed again with a third author. The risk of bias in the included studies was assessed based on the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions [22]. Two authors assessed the bias risk, and disagreements were discussed with a third author.

Statistical analysis
Relative ratios (RRs) with 95% confidence intervals (CIs) were used to assess dichotomous data, and the weighted mean differences (WMDs) with 95% CIs were used to assess continuous data. We extracted hazard ratios (HRs) with 95% CIs from the publications as a relevant measure for the effects of stent patency and patient survival. We estimated the HRs from log-rank Chi 2 statistics, log-rank p values, the given numbers of events, or Kaplan-Meier curves using the methods described by Tierney et al and Parmer et al [23,24]. The Generic Inverse Variance method was applied to analyze time-to-event data.
We performed this meta-analysis following the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions [22], and our report is in accordance with the PRISMA statement (S1 PRISMA Checklist). A fixed-effects model was used if no statistically significant heterogeneity was found; otherwise, a random-effects model was applied. Heterogeneities were analyzed by calculating the Chi 2 and I 2 statistics. I 2 >50% indicated substantial heterogeneity. If significant heterogeneity was found, the potential reasons for the heterogeneity were explored.
All statistical analyses were performed independently and in duplicate by two authors using RevMan 5.2 (Cochrane Collaboration, Oxford, UK) and Stata 11.0 (StataCorp, College Station, Texas, USA).

Search results
The study screening process is shown in Fig 1. A total of 2579 potential abstracts and titles were identified through the systematic search. In all, 832 were duplicates, and a total of 1723 citations were excluded because they were not relevant to the studied topic. Among the remaining 24 citations, 5 trials were not RCTs, 4 were duplicate publications from a conference, and 1 involved handcrafted covered stents. Accordingly, we ultimately selected 14 trials for our meta-analysis.

Risk of bias
We conducted a qualitative risk assessment for each trial. Of the 14 studies included in this meta-analysis, the risk of bias was considered unclear for the abstracts, whereas the full-text trials were assessed in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (S1 and S2 Figs).

Secondary outcomes
Stent dysfunction. Ten studies involving 1225 participants reported stent dysfunction affecting 154 patients with UCSEMSs and 142 patients with CSEMSs. For overall stent dysfunction, the studies showed no significant difference between the two groups, with significant between-study heterogeneity, as shown in In contrast, the rate of tumor ingrowth in the UCSEMS group was significantly higher than the rate in the CSEMS group (RR 0.25, M-H, random, 95% CI 0.12-0.52; p = 0.002, I 2 = 40%) (Fig 5).

Discussion
Patients with inoperable malignant distal biliary obstruction have poor quality of life and bad prognoses. The remarkable advantages of SEMS displacement include longer stent patency and fewer repeat insertions compared with plastic stents [33,34]. Nonetheless, stent dysfunctions caused by sludge formation and tumor ingrowth are the main problems affecting the use of UCSEMSs, whereas CSEMSs were designed to avoid tumor ingrowth. However, whether CSEMSs or UCSEMSs provide better therapeutic efficacy remains unclear. Thus, we performed this review to investigate whether the efficacies of the two types of stents differ significantly.
Our meta-analysis included 14 eligible prospective RCTs that compared cumulative stent patency, overall patient survival, and complication rates between UCSEMSs and CSEMSs. Of the 14 trials, 7 were published as abstracts, and the risks of bias for these trials are unknown. The sensitivity analysis excluded the 7 abstracts and remained consistent with the pooled trials. The 3 RCTs [25,28,30] concluded that UCSEMSs exhibited a significant longer mean patency than CSEMSs did. An opposite conclusion on stent patency was reported by Kitano et al and Krokidis et al [14,15]. Our meta-analysis revealed no significant difference between the two groups, which is consistent with the remaining studies. However, a different conclusion regarding stent patency was reached by two previous meta-analyses. One study concluded that when CSEMSs were applied, they provided significantly longer stent patency than UCSEMSs did [20] (WMD 60.56 days; 95% CI 25.96-95.17; p = 0.001; 3 studies; I 2 = 0%). In contrast, the other study suggested that the two stents did not significantly differ in the proportions of stent patency at 6 (OR 1.82; 95% CI 0.63-5.25) and 12 (OR 1.25; 95% CI 0.65-2.39) months [21]. However, the WMD in stent patency favored the CSEMSs (67.9 days 95% CI 60.3-75.5), presumably because only two trial data points could be merged. WMDs or odds ratios (ORs) were applied in the two previous meta-analyses [20,21] to assess the duration and proportion of stent patency. The data were reported as means, ranges, medians, interquartile ranges or first quartiles, which limited data consolidation and therefore prevented the use of a summary value. These limitations may be why the two meta-analyses failed to reach a unanimous conclusion. Moreover, stent patency and patient survival are time-to-event outcomes; notably, ORs only measure the number of events, which is less appropriate for analyzing time-to-event outcomes. The most appropriate analytical approach for time-to-event outcomes is the HR, which considers the number and timing of events [24]. Several studies [12-16, 25, 26] have reported that using cumulative stent patency is more suitable for comparing stent patency between two groups, and thus, we considered this measure in our meta-analysis. In particular, we extracted HRs from a Kaplan-Meier graph of cumulative stent patency. However, a summary of the HRs suggests no significant difference in cumulative stent patency (HR 0.93, 95% CI 0.19-4.53; p = 0.93), without significant between-study heterogeneity (Chi 2 = 0.31, df = 6, I 2 = 0%).
Patient survival mainly depends on the tumor type, metastatic rate and eventual comorbidities. We pooled the data from the selected studies and found no differences in the etiological factors and tumor metastatic rates between the two groups. Three studies [12,16,26] reported tumor metastases in 36% of patients in the UCSEMS group and 42% of patients in the CSEMS group. However, we did not address the eventual comorbidities in the two groups. In our metaanalysis, only one study [15] suggested a significant difference in patient survival between the UCSEMS and the CSEMS groups (median 180.5 vs 243.5; p = 0.039). In contrast, the remaining studies suggested no difference in overall survival between the two groups. Kitano et al [14] also examined the patient survival time without stent dysfunction, and they recorded a longer median time when CSEMSs were used compared with when UCSEMSs were used (median 187 vs 132 days; p = 0.043).
Although using SEMSs prolongs stent patency, stent dysfunction accounts for occlusion is the main problem associated with SEMSs. Stent dysfunction occurs for all types of SEMSs, but the mechanisms of different stent types differ. Sludge formation, tumor ingrowth, tumor overgrowth and stent migration are the main reasons for stent dysfunction. Although tumor ingrowth may occur in a covered stent segment because of covering membrane fissuring [35],  several trials confirmed that UCSEMSs were more prone to biliary re-obstruction because of tumor ingrowth. Our review supports UCSEMSs exhibiting a higher rate of tumor ingrowth.
The SEMS coating membrane is intended to reduce recurrent biliary obstruction because of tumor ingrowth. Using covered stents has been shown to reduce the rate of tumor ingrowth, as stated above. However, our review shows that CSEMSs exhibit higher rates of sludge formation, tumor overgrowth and stent migration than UCSEMSs do. A previous study suggested that patients with CSEMSs show a higher rate of sludge formation than patients with UCSEMSs do (4-7% vs 0-3%) [14]. We speculated that the covering membrane favors microbial accumulation and thus is susceptible to bacterial biofilm development, as observed for plastic stents [5,33]. Moreover, long stents with a high axial force kink easily, which slows biliary drainage and facilitates sludge formation. Because of the higher rate of migration observed in the CSEMS group, the coating membrane may prevent stent embedding, thus inhibiting stent integration into the bile duct wall. Moreover, stents with a high axial force can increase the risk of migration and bile duct kinking [36]. Interestingly, overall stent dysfunction did not differ significantly between CSEMSs and UCSEMSs. Two possible explanations for this observation are as follows. First, the benefits of preventing tumor ingrowth are offset by the adverse events caused by the covering membrane. Second, our meta-analysis only included the four main reasons (sludge formation, tumor ingrowth, tumor overgrowth and stent migration) for overall stent dysfunction; other reasons, such as food debris, were ignored because of incomplete data, although we attempted to contact the study authors about these data. Meta-regression was performed to explore the sources of heterogeneity existing in overall stent dysfunction and tumor ingrowth outcomes. However, data were presented in various formats in different studies, and incomplete data limited data consolidation. In this meta-analysis, the stent placement approach, the number of endoscopic sphincterotomy the research center and the region were regarded as variables in the meta-regression. Unfortunately, the four variables did not seem to contribute to the heterogeneity in the two outcomes cited above (S3 and S4 Figs). Rather, the heterogeneity might have resulted from the other clinical conditions, such as the stent length and stent diameter.
As stated above, our meta-analysis indicates that there is no significant difference in primary stent patency, overall stent dysfunction between two groups. Thus, there is a misunderstanding that the efficiency of the two stents for managing malignant distal biliary obstruction is equivalent. However, the fact is that the follow-up period in these RCTs only lasted from primary stents insertion to stents dysfunction or patients death, but not taken the further management for patients with occluded stent into account. The intervals and the rate of overall stent dysfunction were similar between two groups which also mean the ratio of patients who need further management is similar too. Kida M, et al [37]reported that nearly 50% survival patients with SEMSs had a stent occlusion, and the patency duration of further management with a balloon mechanical cleaning or "stent-in-stent" placement was rather short. Besides, removal of the occluded stents is better for preserving patient quality of life [37]. However, removal of UCSEMSs is more risky and tough than removal of CSEMSs because of tumor ingrowth [38,39]. Familiari P, et al [40] reported that 0% of UCSEMSs and 86.4% of CSEMSs could be removed without serious complications after stents occlusion. A subsequent study [41] suggested that 38.4% occluded stent of UCSEMs was able to be removed, but the ratio was far lower than 92.3% of CSEMSs. Under this condition, in order to simplify the further management and preserve patient quality of life, CSEMSs is a better choice for patients with malignant biliary obstruction due to their removability.
Several investigators have reported that pancreatitis is the most common adverse event after SEMS placement, with an incidence range of 0-8.8% [34,[42][43][44][45]. Mechanical injuries from pancreatic duct manipulation and pancreatic orifice occlusion are the two major etiological causes of post-SEMS-placement pancreatitis. Moreover, SEMSs with a high axial force may increase the pancreatitis incidence compared with plastic stents. According to previous studies [1,25,[46][47][48], CSEMSs might increase the incidence of pancreatitis and cholecystitis by covering the pancreatic duct orifice or cystic duct. We collected data on this topic from all but four of the selected trials [13,27,28,31]; the excluded trials did not provide exact numbers. We found that of 464 patients with UCSEMSs, 24 suffered from cholecystitis, and 6 suffered from pancreatitis. Of the 472 patients who had CSEMSs, 25 suffered from cholecystitis, and 9 suffered from pancreatitis. Thus, we cannot conclude that patients with CSEMSs are prone to pancreatitis and cholecystitis. Moreover, none of the trials reported the pancreatic duct conditions or whether the pancreatic exocrine function remained normal. A distal bile duct that is obstructed or infiltrated by a tumor before SEMS placement may explain why CSEMSs near the papilla of Vater do not necessarily result in acute pancreatitis [16]. We also analyzed the overall complication rate, unexpectedly; CSEMSs did not reduce the overall complication rate after stent insertion compared with UCSEMSs. Additionally, we analyzed the complication classification between the two groups. The trials [15,16,25] classified the complications from minor to serious, in accordance with the Society of Interventional Radiology classification system for complications. Telford et al [13] defined a serious complication as one requiring an invasive procedure or hospitalization or resulting in death. The means used to grade the complications in the remaining studies were unclear. Interestingly, the overall complication, minor complication and serious complication rates did not differ between the two groups in only four trials [12,[14][15][16]. However, when we considered an additional trial [13], CSEMSs exhibited a higher rate of serious complications compared with UCSEMSs. Therefore, because the definition of a serious complication was inconsistent, different conclusions were reported.
Regarding biliary drainage efficiency, changes in serum bilirubin can be observed after stent insertion. Hyperbilirubinemia is a type of liver disorder that can produce biliary cirrhosis if the obstruction cannot be resolved for a long period of time. Many patients cannot bear the pruritus caused by high-level bilirubin. For patients undergoing palliative treatment, biliary drainage plays a fundamental role in improving quality of life. To assess the biliary drainage efficacy, we compared the changes in serum bilirubin at 2 weeks after stent insertion between the two groups. Only two studies [14,25] reported serum bilirubin values, and our review suggested no significant difference between UCSEMSs and CSEMSs. The change in serum bilirubin at 2 weeks after stent placement can be used as a short-term evaluation of biliary drainage. In the present analysis, the success rate of stent deployment did not vary, which is the main reason for the equivalent biliary drainage between the two groups, and the biliary drainage did not differ.
Certain limitations of our meta-analysis should be noted. First, although we included many studies, we did not investigate the potential publication bias of the abstracts included in this review. Second, although the patient characteristics (age and gender) did not differ, important information, such as eventual comorbidities, stent length, axial force and the stent placement approach, could not be pooled, which might have affected the conclusions regarding clinical efficacy.

Conclusions
Our meta-analysis indicates that patients with malignant distal biliary obstructions treated with CSEMSs do not experience an extra benefit compared with patients treated with UCSEMSs during the period from primary stent insertion to primary stent dysfunction or patient death. However, given the further management for occluded stent, we believe that using CSEMSs is a better choice for patient undergoing palliative treatment for malignant distal biliary obstruction.