Introduction

Delayed gastric emptying (DGE) has been one of the common, yet unresolved, postoperative complications after pancreaticoduodenectomy (PD) operations performed for the treatment of pancreatic head cancer.1 – 3 Postoperative DGE can significantly affect the quality of life in this already disadvantaged population, where only 10 to 15 % of pancreatic head cancer patients are candidates for surgical resection upon presentation with a dismal 5-year survival of 5 % of all comers.4 – 6 The mechanism of DGE is still unclear but is suggested to be due to the extent of gastric resection, loss of the pylorus, interrupted gastrointestinal neural connections, diabetes, local ischemia, or loss of gastrointestinal hormonal production causing gastroparesis.7 , 8 In addition, some postoperative complications have been suggested to also affect the incidence of DGE, including pancreatic fistula and abscesses as well as the method of reconstruction of the gastrointestinal continuity.9

The classic pancreaticoduodenectomy (cPD) described by Whipple included distal hemigastrectomy.1 It had a reported incidence ranging 20–40 % of DGE.10 – 12 Subsequently, new alternative surgical techniques were suggested to attempt to decrease the incidence of DGE. Pylorus-preserving pancreaticoduodenectomy (PPPD) was developed by Traverso and Longmire, and it included full preservation of the pylorus with its nerve supply.13 However, there has been no evidence from prospective studies and meta-analyses to indicate the superiority of PPPD in terms of DGE, long-term results or quality of life.14 – 16 Therefore, more recently, and mostly performed in Japan since the late 1990s, subtotal stomach-preserving pancreaticoduodenectomy (SSPPD) in which the pyloric ring only is removed with preservation of about 90 % of the stomach has been described.17 , 18 In that procedure, the gastric outlet is thought to be wider than that in PPPD; together with the preservation of the blood supply and the innervation of the prepyloric region, it is thought to may help reduce the incidence of DGE.19 – 22

Many studies published contradicting results as to the incidence of DGE in patients who had SSPPD as compared to those who had PPPD.17 , 19 , 23 – 28 Hence, the aim of this study is to review and pool the data from published literature comparing the incidence of DGE between these two techniques.

Materials and Methods

Literature Search and Study Selection

A comprehensive search of MEDLINE, EMBASE, Google Scholar, Scopus, and the Cochrane database was performed for all articles published in the English language comparing the outcomes including DGE after SSPPD vs. PPPD. The search was conducted using the following MeSH terms: “pancreaticoduodenectomy with delayed gastric emptying,” “subtotal stomach preserving pancreaticoduodenectomy AND delayed gastric emptying,” “pylorus preserving pancreaticoduodenectomy with delayed gastric emptying,” and gastrojejunostomy. The related-articles function was used to expand the search from each relevant study identified. All citations and abstracts identified were thoroughly reviewed. The latest search was performed on October 20, 2014. Bibliography of retrieved papers was further screened for any additional eligible studies.

Outcomes of Interest

Only studies reporting on the comparison between SSPPD vs. PPPD were included. The primary end point of the study was DGE. The secondary end points included patient demographics, operative blood loss, operative time and other outcomes like pancreatic fistula, and mortality.

Inclusion Criteria

In order to be included in the analysis, studies had to

  1. 1.

    Compare the outcome measures mentioned above between patients who had SSPPD and those who had PPPD

  2. 2.

    Report on at least one of the outcomes of interest mentioned above

When the same institution reported two studies, we included either the one of better quality (e.g., larger sample size) or the most recent publication.

Exclusion Criteria

Studies were excluded from analysis if

  1. 1.

    They were either noncomparative studies or case series.

  2. 2.

    The outcomes of interest were not reported for the two techniques.

  3. 3.

    There was an overlap between authors, institutions, or patient cohorts.

Definitions

DGE was evaluated mostly as defined and graded by the International Study Group of Pancreatic Surgery (ISGPS) classification.6 , 29

  • Grade A, nasogastric tube (NGT) required for more than three postoperative days (POD) or the inability to tolerate a solid diet by POD 7

  • Grade B, NGT required for 8 to 14 days, the need for reinsertion of the NGT tube after 7 days, or the inability to tolerate a solid diet by POD 14

  • Grade C, NGT lasting more than 14 days, the need for NGT reinsertion after 14 days, or the inability to tolerate a solid diet by POD 21

In the remaining two studies,17 , 19 DGE was defined as NGT use ≥10 days or solid food tolerance ≥14 days postoperatively.

Surgical Techniques

  • SSPPD involved a division of the stomach at about 2 cm proximal to the pyloric ring.

  • PPPD involved division of the proximal duodenum at about 3 cm distal to the pyloric ring.

The rest of the resection for all patients was done as described in the original cPD1 (Fig. 1).

Fig. 1
figure 1

Schematic illustration of the three types of PD: conventional PD with cPD, SSPD, and PPPD. In SSPD, the stomach is resected at 2–3 cm proximal to the pyloric ring

Full size image

Reconstruction for all patients was performed using a modified Child’s method consisting of an end‐to‐side pancreatojejunostomy, an end‐to‐side choledochojejunostomy, and an end-to-side antecolic gastrojejunostomy.30

Data Extraction and Quality Assessment

Two reviewers (M.H. and R.G.) independently extracted the following data from each study: study characteristics (first author, year of publication, study design) and population characteristics (number of patients included, age, sex, and tumor characteristics). Where available, DGE, pancreatic fistula, abdominal collection/abscess, operative blood loss, operative time, and overall morbidity and mortality were recorded for each study. Quality assessment of the nonrandomized cohort studies was performed using the 22-item Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist; median score was 27. Quality assessment of the prospective studies was performed using the 25-item Consolidated Standards of Reporting Trials (CONSORT) checklist; median score was 26. The observational studies were mainly deficient in indicating the study design in the title or abstract, description of sensitivity analyses, and translation of relative risk estimates into absolute risk. Explanation on handling of missing data, use of a flow diagram to indicate study participants, and reporting of unadjusted, confounder-adjusted estimates were lacking in three studies. Lastly, all but one study did not describe any efforts to address the potential sources of bias. The randomized controlled trials were deficient in reporting the methods of allocation concealment, methodology used for blinding, and lastly registration of the trial protocol.

Statistical Analysis

All statistical analyses were performed using Stata 12 (StataCorp LP, College Station, TX). We analyzed all the dichotomous variables by calculating the risk ratio (RR) and corresponding 95 % confidence interval (CI) as the summary statistic. For continuous outcomes, we pooled the data using the weighted mean difference (WMD) and converted continuous data that was reported as median and range to the mean and standard deviation. All p values of <0.05 (two tailed) were considered statistically significant. For the meta-analysis, we used both the fixed-effects and random-effects model (weighted by the Mantel-Haenszel method),31 with using the random-effects model in the presence of clinical or statistical heterogeneity. Clinical heterogeneity between studies was considered when differences in operative technique or definition of DGE were present. Statistical heterogeneity across studies was quantified using the X 2 (or Cochran Q statistic) and I 2 statistic. The I 2 statistic was derived from the Q statistic ([Q − df/Q] × 100), which provides a measure of the proportion of the overall variation attributable to heterogeneity between the studies. Homogeneity was considered absent if the Q statistic showed p < 0.20, and the heterogeneity was considered significant when the I 2 statistic exceeded 50 %. Subgroup analyses were done for patients with postoperative complications developing DGE.

Results

Literature Search

Our search yielded 233 publications. We excluded 181 studies at the title and abstract levels, selecting 52 studies for full-text review. Out of the 52 selected for full-text review, 44 were excluded based on inclusion and exclusion criteria. Thus, we identified 5 cohort studies, 2 randomized clinical trials, and 1 nonrandomized prospective clinical trial with a total of 663 patients for inclusion in the meta-analysis (Fig. 2)17 , 19 , 23 – 28; 309 (47 %) patients underwent PPPD and 354 (53 %) patients underwent SSPPD. Median age was 65 years (IQR 61–67) with an average male/female ratio of 57 vs. 43 %. There was no difference between the two groups regarding age, sex, histopathology, and preoperative comorbidities (Table 1).

Fig. 2
figure 2

Systematic literature search and selection strategy

Full size image
Table 1 Demographic characteristics of included studies
Full size table

Primary Outcome

All eight studies included assessable data on postoperative incidence of DGE. In all but two studies, DGE was defined using the ISGPS grading criteria23 – 28; the two remaining studies defined DGE as NGT ≥ POD 10, differing on the intake of solid food vs. regular diet at POD ≥ 1417 , 19 (Table 2). When all studies were included, SSPPD resulted in a lower incidence of DGE (SSPPD vs. PPPD, 21 vs. 37 % (RR 0.527; 95 % CI 0.363–0.763; p < 0.001)). There was moderate heterogeneity (I 2 = 52.1 %; p < 0.04), mainly explained by one study that had a RR of 1.13 (Fig. 3). Subgroup analysis of studies that strictly used ISGPS definition for DGE showed consistent lower incidence of DGE with SSPPD (SSPPD vs. PPPD, 23 vs. 40 %; RR 0.528; 95 % CI 0.353–0.790; p = 0.002) (Fig. 4). Subgroup analysis of patients who had DGE grade B (SSPPD vs. PPPD, 6 vs. 16 %; RR 0.395; 95 % CI 0.232–0.673; p = 0.001) (Fig. 5) and DGE grade C (SSPPD vs. PPPD, 9 vs. 26 %; RR 0.349; 95 % CI 0.164–0.739; p = 0.006) (Fig. 6) showed that it was significantly less in SSPPD.

Table 2 Postoperative outcomes of the included studies
Full size table
Fig. 3
figure 3

Forest plot for risk of developing delayed gastric emptying with subtotal stomach-preserving pancreaticoduodenectomy (SSPD) vs. pylorus-preserving pancreaticoduodenectomy (PPPD). D + L DerSimonian-Laird random-effects model; M – H Mantel-Haenszel fixed-effects model; RR risk ratio; CI confidence interval

Full size image
Fig. 4
figure 4

Forest plot for risk of developing delayed gastric emptying with subtotal stomach-preserving pancreaticoduodenectomy (SSPD) vs. pylorus-preserving pancreaticoduodenectomy (PPPD) limited to studies using ISGPS definition for DGE only. D + L DerSimonian-Laird random-effects model; M – H Mantel-Haenszel fixed-effects model; RR risk ratio; CI confidence interval

Full size image
Fig. 5
figure 5

Forest plot for risk of developing delayed gastric emptying (grade B) with subtotal stomach-preserving pancreaticoduodenectomy (SSPD) vs. pylorus-preserving pancreaticoduodenectomy (PPPD). D + L DerSimonian-Laird random-effects model; M – H Mantel-Haenszel fixed-effects model; RR risk ratio; CI confidence interval

Full size image
Fig. 6
figure 6

Forest plot for risk of developing delayed gastric emptying (grade C) with subtotal stomach-preserving pancreaticoduodenectomy (SSPD) vs. pylorus-preserving pancreaticoduodenectomy (PPPD). D + L DerSimonian-Laird random-effects model; M – H Mantel-Haenszel fixed-effects model; RR risk ratio; CI confidence interval

Full size image

Five studies reported the number of days NGT was used.17 , 19 , 24 – 26 In a random-effects model, the meta-analysis showed less postoperative NGT days with SSPPD (RR −0.544; 95 % CI −876 to −0.008; p = 0.047) (Fig. 7). There was a moderate heterogeneity amongst the included studies (τ 2 = 23.37; df = 4; p < 0.001; I 2 = 82.9 %).

Fig. 7
figure 7

Forest plot comparing nasogastric tube duration (days) with subtotal stomach-preserving pancreaticoduodenectomy (SSPD) vs. pylorus-preserving pancreaticoduodenectomy (PPPD). D + L DerSimonian-Laird random-effects model; M – H Mantel-Haenszel fixed-effects model; RR risk ratio; CI confidence interval

Full size image

Secondary Outcomes

Overall morbidity was 35 vs. 31 % and postoperative mortality was 0.3 vs. 1.1 %, respectively, for SSPPD vs. PPPD. Seven studies reported the operative blood loss,17 , 18 , 24 – 28 PPPD patients had significantly lower operative blood loss (RR 0.335; 95 % CI 0.096–0.575; p = 0.006) (Fig. 8), and there was moderate heterogeneity amongst the included studies (τ 2 = 11.26; df = 6; p = 0.081; I 2 = 46.7 %). There was no significant difference between the two groups in operative time (RR 0.133; 95 % CI −0.168 to 0.433; p = 0.387). There was a moderate heterogeneity amongst the included studies (τ 2 = 17.77; df = 6, p = 0.007, I 2 = 66.2 %). There was no significant difference in length of hospital stay between the two groups (RR −0.199; 95 % CI −0.516 to 0.119; p = 0.220). There was a moderate heterogeneity amongst the included studies (τ 2 = 19.76; df = 6; p = 0.003; I 2 = 69.6 %). There was no association between operative type and pancreatic fistula (SSPPD vs. PPPD, 20 vs. 21 %; RR 0.901; 95 % CI 0.670–1.211; p = 0.49) and operative type and abscess collection (SSPPD vs. PPPD, 5 vs. 6 %; RR 0.905; 95 % CI 0.483–1.697; p = 0.76).

Fig. 8
figure 8

Forest plot comparing operative blood loss (ml) with subtotal stomach-preserving pancreaticoduodenectomy (SSPD) vs. pylorus-preserving pancreaticoduodenectomy (PPPD). D + L DerSimonian-Laird random-effects model; M – H Mantel-Haenszel fixed-effects model; RR risk ratio; CI confidence interval

Full size image

Discussion

SSPPD has been recently used in surgery for pancreatic head resection for different indications by removing the pyloric ring only with preserving more stomach than in cPD in an attempt to lower the incidence of DGE which has not been proven as a benefit of PPPD. Meta-analysis of eight included studies in our review showed that the incidence of DGE is lower in SSPPD as compared to PPPD with shorter duration of NGT use.

The pathogenesis of the DGE is not clear yet, but it appears to be multifactorial, in spite of the suggested cause of DGE that it may be due to devascularization and denervation of the pyloric region.7 , 8 Our results undermine this theory, as SSPPD does not even preserve the pyloric ring as opposed to PPPD. This benefit may then be attributed to a wider gastric outlet used in gastrojejunostomy reconstruction after SSPPD.19 – 22 Nonetheless, this cannot be the only factor as cPD with hemigastrectomy has even a wider gastric outlet than SSPPD, yet SSPPD has the advantage of preserving innervation and vascularity to the prepyloric area. As PD is most commonly performed in pancreatic cancer which is well known for its biologic aggression, some authors are more of proponents of resection of the pyloric ring to allow for proper skeletonization of the hepatoduodenal ligament for radical lymph node dissection. SSPPD allows for that and, as shown in our analysis, has less incidence of DGE.

DGE was significantly less in SSPPD even in grade B/C ISGPS definition, which is the rather clinically relevant result as grade A DGE usually fairs well clinically with an acceptable postoperative course. In spite of the fact that two of the studies17 , 19 predated the ISGPS definition of DGE, those two studies included detailed criteria of defining DGE and were similar in that, allowing including them in the overall analysis, this result was also reiterated by subgrouping those studies that strictly used the ISGPS criteria.23 – 28

Postoperative complications like pancreatic fistula and abdominal collection/abscess were reported in some studies to be risk factors for DGE,9 our analysis showed that there was no statistically significant difference between the groups regarding these complications. Despite there being no reported data in the included studies to detail whether patients who developed these complications were the ones to develop DGE, the lack of difference in these complications’ incidence between groups supports that the operative technique may have been the main variable associated with decreased incidence of DGE. Another factor, preoperative diabetes mellitus (DM), was reported to be a risk factor for the DGE9 , 29; only four studies in our analysis reported on it and it was not found to have a significant effect on the incidence of DGE.

Limitations of this study include relatively small sample sizes of included studies in light of such a multifactorial outcome in question. The studies included in our analysis were of acceptable quality, the nonrandomized cohort studies had a STROBE median score of 27, and the prospective studies had a CONSORT median score of 26; both factors underline the validity of the analyzed data. Five of the eight studies included were retrospective, with above-average heterogeneity at times. In the included studies, there were limited subgroup analyses to try to explain the association of DGE with other complications. Finally, technical heterogeneity between surgeons and centers is always a factor in these comparisons.

Conclusion

SSPPD was found to have lower incidence of DGE than PPPD. In their current design, the included analysis might not completely elucidate the correlation of DGE with other perioperative complications. Therefore, standardized randomized prospective studies may be helpful to investigate whether DGE is associated with other risk factors and postoperative complications, or it is a complication resulting from a specific surgical technique.