Transpancreatic Sphincterotomy Is Effective and Safe in Expert Hands on the Short Term

A systematic literature search was conducted to find all relevant articles containing data on TPS in accordance with the PRISMA guideline [13]. The search strategy included the following terms: “transpancreatic septotomy” or “transpancreatic sphincterotomy” or “transpancreatic septostomy” or “transpancreatic precut sphincterotomy” or “pancreatic sphincterotomy” or “transpancreatic papillary septotomy” or “transpancreatic sphincter precut” or “transpancreatic duct precut” or “pancreatic sphincter precutting” or “pancreatic precut sphincterotomy” or “transpancreatic precut septotomy” or “transpancreatic precut septostomy” or “pancreatic septotomy” or “pancreatic septostomy” or “pancreatic precut” or “transpancreatic precut” or “transpancreatic.” EMBASE, PubMed, Scopus, Web of Science, ProQuest, and Cochrane Library databases were searched from their inception till February 8, 2018.

In order to compare TPS to DGW and NKPP, only prospective studies were included. However, only retrospective data were available in the comparison of TPS–NKF, and these were also included in our analysis. Appropriate conference abstracts were also analyzed to minimize publication bias, and additional subgroup analyses excluding them were carried out to show their effects on outcomes.

Comparative and also non-comparative prospective and retrospective studies were included in the calculation of overall success and complications rate of TPS. Randomized controlled trials (RCT) and prospective and retrospective observational studies were analyzed separately (Table 4).

Titles and abstracts of studies identified were screened by two authors (D.P. and Á.V.) independently, and then, the full-text articles were searched to identify eligible studies. Data extraction and risk of bias assessment were done independently by the authors. Peer-reviewed works and conference abstracts were included. Unpublished data were not requested from the authors. Any disagreement was resolved by discussion in plenum. Prophylactic measures to prevent PEP; furthermore, the length and results of follow-up were also collected and analyzed.

The Newcastle–Ottawa scale (NOS) was used for prospective and retrospective studies to assess risk of bias within the individual studies [14] (Table 5). Randomized controlled trials were assessed by the Cochrane Risk of Bias Tool [15] (Table 6).

Pooled odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated to compare the biliary cannulation success and PEP rates among the different cannulation techniques. Risk difference (RD) was calculated to compare the bleeding and perforation rates in order to avoid overestimation since OR or RR calculations would exclude those studies where zero events were reported. The random-effect model of DerSimonian and Laird [16] was used in meta-analysis. Subgroup analyses excluding studies with sequential designs and that reported only in an abstract format were also carried out. Sensitivity analyses were carried out using four types of summary statistics (RR [risk ratio] vs. OR vs. RD vs. Peto’s OR) and two types of meta-analytical models (fixed vs. random effects) to test the robustness of our findings [17]. Heterogeneity was tested with two methods, namely the Cochrane’s Q and the I² statistics. The Q test was computed by summing the squared deviations of each study’s estimate from the overall meta-analysis estimate; P values were obtained by comparing the statistical results with a χ² distribution with k − 1 degrees of freedom (where k was the number of studies). A P value of less than 0.1 was considered suggestive of significant heterogeneity. The I² statistic represents the percentage of the total variability across studies that is due to heterogeneity, i.e., I² value between 0 and 40% indicates low, 30–60% moderate, 50–90% substantial, and 75–100% considerable heterogeneity, based on Cochrane Handbook for Systematic Reviews of Interventions [17]. Publication bias was planned to be examined by visual inspection of funnel plots and the Egger’s method [18]. Meta-analytical calculations were done with Review Manager (RevMan) computer program (version 5.3, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).

Altogether, 2787 records were identified during database search: 510 in EMBASE, 339 in PubMed, 968 in Scopus, 255 in Web of Science, 544 in ProQuest, and 171 in Cochrane Library, respectively. The latest search was run on February 8, 2018, and finally, 33 relevant studies were included in the qualitative synthesis, while data from 14 studies were extracted for the meta-analysis (Fig. 1).

Characteristics of the included studies with the applied PEP prophylaxis (Table 1), the definitions of difficult biliary access and the endoscopists/centers experience (Table 2), and the late adverse events are summarized in Table 3.

Three RCTs [19‐21] and two prospective observational studies [22, 23] compared TPS and DGW. One of them was only available in abstract form [19]. Two of them used a sequential design [22, 23], applying TPS only after DGW, as a rescue technique.

Two RCTs [24, 25] and three prospective observational studies [22, 23, 26] provided data on the comparison of TPS vs. NKPP, two of them with sequential design [22, 23]. New prospective studies were not identified after our previous meta-analysis; however, we conducted further sensitivity and subgroup analyses in this comparison [6].

Comparison of TPS and NKF was not found in any prospective studies; four retrospective studies (two of them only in abstract form) were identified and analyzed to synthesize available comparative evidence [9, 27‐29].

Two prospective case series of TPS without relevant comparisons to other advanced cannulation methods [30, 31] and, additionally, 23 retrospective observational studies with reported outcome data were included in the pooled analyses of overall outcomes of TPS [4, 9, 10, 27‐29, 32‐48] (Table 4).

The risk of bias in the prospective (not RCTs) and the four retrospective studies included in the meta-analyses was analyzed with the NOS (Table 5). In most of the full-text studies, baseline characteristics of cohorts were reported with comparable, homogeneous groups. Technical details of interventions were thoroughly reported; all full-text studies defined precut methods appropriately. On the other hand, definitions of adverse outcomes somewhat varied; however, most studies used the consensus definitions [49]. The appropriate length of follow-up is questionable in the cases of late adverse events, and only one prospective study reported the length of follow-up as longer than 30 days [30]. The abstracts contained limited information about the above-mentioned details; therefore, they carry an unclear risk of bias.

In case of RCTs, the Cochrane Risk of Bias Tool was used (Table 6). Only one study [21] reported the method of randomization and the method of ensuring allocation concealment. Blinding in studies of endoscopic interventions at participant and personnel level is difficult to execute and therefore could not be expected. However, blinded late outcome assessment (PEP, late bleeding, perforation) could be arranged more easily. Nevertheless, none of the studies reported blinding of any kind. Three out of five RCTs did not report the rate of cholangitis; therefore, this outcome could not be analyzed [19, 24, 25]. One RCT was only published in abstract form which makes the data quality questionable; consequently, this study carries a high risk of bias [19].

Publication bias could not be reliably assessed based on funnel plots or by the Egger’s method because of the small number of included studies. According to the Cochrane Handbook, funnel plots and other statistical tests are not advised to assess small study effect and publication bias under ten studies per analysis [17, 18, 50].

Most of the prospective studies reported endoscopists’ experience in yearly case numbers, some in lifetime ERCP numbers, too. Based on the reported numbers, all endoscopists performed more than 200 ERCPs/year. In one study, the caseload of the endoscopists exceeded 500 ERCPs annually [30]. Trainee participation was not reported in any of the studies. Most of the centers reported high-volume ERCPs (even above 1000 procedures/year [23, 24]), only one study [9] reported lower numbers (< 300 ERCPs/year), while no information was found about center or endoscopist caseload in one study [29] (Table 2).

TPS showed superiority in success rate compared to DGW (OR 2.72; 95% CI 1.30–5.69; 176 and 235 patients, respectively; I² = 50%) (Fig. 2a) and NKPP (OR 2.32; 95% CI 1.37–3.93; 292 and 260 patients, respectively; I² = 7%) (Fig. 2b). The success rate of TPS and NKF did not differ (OR 1.38; 95% CI 0.32–5.96; 295 and 141 patients, respectively; I² = 22%) (Fig. 2c).

In the TPS versus DGW comparison of cannulation success rates, no significant difference was detected between the two methods if only RCTs were included (OR 3.02; 95% CI 0.73–12.59; 113 and 107 patients, respectively; I² = 69%), probably because of the greater confidence intervals of the results. On the other hand, subgroup analysis of full-text studies found the superiority of TPS over DGW with regard to cannulation success rate (Suppl. Figure 1).

The overall success rate of TPS in prospective studies was 89.7% (564/629). The success rate was the same if all studies were analyzed (89.6%, 2343/2615), as well as the separate analysis of RCTs resulted in similarly high value (91.7%, 199/217) (Table 4).

No significant difference was found between the TPS versus DGW (OR 0.72; 95% CI 0.24–2.10; 151 and 134 patients, respectively; I² = 55%) (Fig. 3a) and TPS versus NKPP (OR 1.63; 95% CI 0.48–5.47; 265 and 242 patients, respectively; I² = 57%) (Fig. 3b) comparisons. However, the TPS technique showed a higher PEP rate compared to NKF method (OR 4.62; 95% CI 1.36–15.72; 295 and 141 patients, respectively; I² = 16%) (Fig. 3c).

If we excluded abstracts from the NKF versus TPS comparison, the significant difference disappeared (OR 3.49; 95% CI 0.20–62.21; 86 and 115 patients, respectively; I² = 63%) and expectedly, a wide confidence interval could be seen (Suppl. Figure 2). In the other subgroups, no differences were found when sequential studies or abstracts were omitted from the analyses. Inclusion of RCTs only did not result any change in significance regarding TPS versus DGW and TPS versus NKPP comparisons.

The overall PEP rate of TPS was 8.1% (49/604) in prospective studies, 7.1% (183/2590) in all studies, and 7.4% (16/217) in RCTs (Table 4).

Only one recently published study used PPS in all patients undergoing TPS [20], while all the others reported no or only some PPS implantation in the TPS cases (Table 1). Pharmacologic prevention of PEP was applied in three studies [20, 27, 28]; however, the recommended nonsteroid anti-inflammatory drug (NSAID) suppositories were not used or not reported in any of the studies included in the meta-analyses (Table 1).

The pooled analysis did not show any difference in bleeding rate when TPS was compared to DGW (risk difference [RD] 0.01; 95% CI − 0.03 to 0.05; 109 and 95 patients, respectively; I² = 0%) (Fig. 4a), NKPP (RD − 0.00; 95% CI − 0.04 to 0.03; 268 and 239 patients, respectively; I² = 20%) (Fig. 4b), and NKF (RD 0.00; 95% CI − 0.03 to 0.03; 295 and 141 patients, respectively; I² = 0%) (Fig. 4c).

Subgroup analyses did not alter the findings of bleeding rates significantly.

The overall bleeding rate of TPS was 3.4% (19/562) in prospective studies, 2.0% (50/2548) in all studies, and 1.7% (3/175) in RCTs (Table 4).

Perforation rates did not differ when comparing TPS versus DGW (RD − 0.01; 95% CI − 0.04 to 0.03; 109 vs. 95; I² = 0%) (Fig. 5a), TPS versus NKPP (RD − 0.00; 95% CI − 0.02 to 0.01; 267 and 240 patients, respectively; I² = 0%) (Fig. 5b), and TPS versus NKF (RD 0.00; 95% CI − 0.02 to 0.03; 295 and 141 patients, respectively; I² = 0%) (Fig. 5c).

Subgroup analyses did not alter the findings in perforations rates significantly.

The overall perforation rate was 0.5% (3/562) in prospective studies, 0.4% (11/2548) in all studies, while 0% (0/175) in RCTs (Table 4).

Application of other meta-analytical models (fixed-effect vs. random-effect analysis) and summary statistics (OR vs. RR vs. RD vs. Peto’s OR) did not affect the outcomes significantly in the main analyses; thus, our conclusions remain unaltered (Suppl. Table 1).

However, subgroup analyses excluding non-RCTs, sequential trials, and studies only available in an abstract form altered significantly some results (i.e., success rate in TPS vs. DGW and PEP rate in TPS vs. NKF comparisons, respectively) (Suppl. Table 2, Figs. 1 and 2).

Pancreatic duct stricture or chronic pancreatitis could potentially develop after pancreatic sphincterotomy; therefore, a longer follow-up period is needed to detect these adverse outcomes [11]. Small caliber pancreatic stents could rarely cause pancreatic ductal changes in long term (1 month or longer) [51, 52]. Only one prospective study, a case series with 116 patients, reported a median 5-month follow-up (range 2–35) with no late adverse events [30]. Another paper similarly did not report late chronic pancreatitis or ductitis from PPS; no strictures were described during longer, however not specified, follow-up [22] (Table 3). A few retrospective studies also published longer-term results: Miao et al. [45] reported no stricture after 4 months of follow-up period, while Barakat et al. [33] found no late stricture formation after an unknown length of “long-term” follow-up.