Perioperative Drug Treatment in Pancreatic Surgery—A Systematic Review and Meta-Analysis

Introduction: Pancreatic resections for malignant or benign diseases are associated with major morbidity and changes in physiology. To reduce perioperative complications and enhance recovery, many types of perioperative medical management have been introduced. The aim of this study was to provide an evidence-based overview on the best perioperative drug treatment. Methods: The electronic bibliographic databases Medline, Embase, CENTRAL, and Web of Science were systematically searched for randomized controlled trials (RCT) evaluating perioperative drug treatments in pancreatic surgery. The investigated drugs were somatostatin analogues, steroids, pancreatic enzyme replacement therapy (PERT), prokinetic therapy, antidiabetic drugs, and proton pump inhibitors (PPI). Targeted outcomes in each drug category were meta-analyzed. Results: A total of 49 RCT were included. The analysis of somatostatin analogues showed a significantly lower incidence of postoperative pancreatic fistula (POPF) in the somatostatin group compared to the control group (OR 0.58, 95% CI: 0.45 to 0.74). The comparison of glucocorticoids versus placebo showed significantly less POPF in the glucocorticoid group (OR 0.22, 95% CI: 0.07 to 0.77). There was no significant difference in DGE when erythromycin was compared to placebo (OR 0.33, 95% CI: 0.08 to 1.30). The other investigated drug regimens could only be analyzed qualitatively. Conclusion: This systematic review provides a comprehensive overview on perioperative drug treatment in pancreatic surgery. Some often-prescribed perioperative drug treatments lack high quality evidence and further research is needed.

Somatostatin analogues are used to lower the amount of pancreatic juices' secretion and therefore may have an influence on POPF [4]. Glucocorticoids do reduce the inflammatory and stress response, a possible mediator of post-pancreatectomy complications [5]. Erythromycin may has an effect on pyloric relaxation and is used for treatment of DGE [6]. Inhibition of gastric acid production by proton pump inhibitors (PPI) may decrease the rate of anastomotic bleeding [7]. Furthermore, PD, DP, or TP are associated with endocrine and exocrine insufficiency. Oral replacement of lipase and pancreatic amylase is used to improve dietary functions, whereas insulin therapy can be given to treat endocrine Furthermore, for each pooled outcome the certainty of evidence was rated to be very low, low, moderate, or high by using the "Grading of recommendations, assessment, Development and Evaluation" System (GRADE) [13].

Statistical Analysis
Trials investigating the same drug were pooled. A sub-analysis for studies using the official ISGPS criteria was performed where applicable [14]. Furthermore, a sensitivity analysis was performed for DP versus PD where applicable. Odds ratios (OR) and 95%-CI were calculated for dichotomous outcomes using the Mantel-Haenszel (M-H) method. Continuous outcomes were pooled as mean difference and 95%-CI with the inverse variance method. If trials only reported medians or ranges, the methods described by Hozo, 2005 were applied to calculate means and standard deviations (SD) from the values reported [15]. Meta-analyses were performed with program R (Version 4.2.0). Forest plots were used for the graphical presentation of effect estimates. Publication bias was explored by funnel plotting if more than 10 trials were pooled. For all outcomes, a random-effect model was applied to account for methodological and clinical differences. Statistical heterogeneity among the effect estimates of the included trials was evaluated using the I 2 statistic. I 2 less than 25% were considered to indicate low heterogeneity and an I 2 > 75% to indicate high heterogeneity [16].

Literature Search
A total of 34,437 articles were evaluated. A detailed description of the screening process can be seen in the PRISMA Flowchart ( Figure 1). Finally, 49 trials fulfilled the inclusion criteria and were therefore included in the final qualitative and quantitative analysis.

Somatostatin Analogues
A total of 33 randomized controlled trials including 3742 patients reported perioperative outcomes of somatostatin use versus placebo after pancreatic resectio ble 1) . Both DP as well as PD were studied. A minority of patients after tot createctomy were also included in this analysis. There was no difference in baselin acteristics. Risk of bias assessment resulted in low risk of bias in 3 studies, some co in 8 studies, and a high risk in 22 studies. Details can be seen in the Supplementary rials.

Somatostatin Analogues
A total of 33 randomized controlled trials including 3742 patients reported on the perioperative outcomes of somatostatin use versus placebo after pancreatic resections (Table 1) . Both DP as well as PD were studied. A minority of patients after total pancreatectomy were also included in this analysis. There was no difference in baseline characteristics. Risk of bias assessment resulted in low risk of bias in 3 studies, some concerns in 8 studies, and a high risk in 22 studies. Details can be seen in the Supplementary Materials. There was no significant difference in mortality in the somatostatin group versus the control group (OR 1.09, 95% CI 0.71 to 1.69, I 2 = 0%, Figure 2). The certainty of evidence was moderate. Incidence of POPF was significantly lower in the somatostatin group compared to the control group (OR 0.58, 95% CI: 0.45 to 0.74, I 2 = 20%, Figure 3). However, the certainty of evidence was low. This effect was consistent with a sub-analysis of POPF (Grade B and C) in studies that used the ISGPS criteria (OR 0.48, 95% CI: 0.31-0.73, I 2 = 0%) [17,23,25,27,34,45]. The sub-analysis included six studies with a total of 833 patients and the certainty of evidence was low. Sensitivity analysis showed a lower rate of POPF in the somatostatin analogue group compared to the control group for both, PD (1842 patients, OR 0.67, 95% CI: 0.52-0.86, I 2 = 1%, Supplementary Materials) and DP (199 patients, OR 0.35, 95% CI:0.14 to 0.86, I 2 = 3%, Supplementary Materials). Overall complications failed to show a significant difference (OR 0.76, 95% CI: 0.56 to 1.02, I 2 = 49%). The certainty of evidence was again low.

Glucocorticoids
Two trials including a total of 93 patients were included in this analysis comparing glucocorticoids versus placebo. The trial of Laainen et al. analyzed high-risk patients after PD, whereas the trial of Antila et al. analyzed patients after distal pancreatectomies [51,52]. Baseline characteristics and treatment regimens are described in Table 2.

Glucocorticoids
Two trials including a total of 93 patients were included in this analysis comparing glucocorticoids versus placebo. The trial of Laainen et al. analyzed high-risk patients after PD, whereas the trial of Antila et al. analyzed patients after distal pancreatectomies [51,52]. Baseline characteristics and treatment regimens are described in Table 2 The 90-day mortality was reported in both trials without any incidents. Pooled analysis of POPF showed significantly less events in the glucocorticoid group when compared to placebo (OR 0.22, 95% CI: 0.07 to 0.77, I 2 = 3%). There was no significant difference in PPH (OR 0.54, 95% CI: 0.14 to 2.03, I 2 = NA), DGE (OR 0.54, 95% CI: 0.2 to 1.45, I 2 = 0%) or intraabdominal fluid collections (OR 0.47, 95% CI: 0.15 to 1.52, I 2 = 0%). LOS was equal in both groups (MD −0.63 days, 95% CI: −3.93 to 2.67, I 2 = 0%). The certainty of evidence of all pooled outcomes of this analysis was very low.

Prokinetics
Two trials with a total of 146 patients were included for this analysis [6,53]. The trial of Ohwada et al. analyzed 31 patients after pylorus-preserving pancreatoduodenectomy [53]. Yeo et al. included a fast majority of pylorus-preserving duodenectomies (85%), but classic Whipple procedures and total pancreatectomies were also analyzed [6]. Treatment regimens and baseline characteristics are presented in Table 3. Overall risk of bias was high in the trial of Yeo et al. and some concerns were seen in the trial of Ohwada et al. Details on risk of bias assessment can be found in the Supplementary Materials.   Both trials reported on DGE and nasogastric tube removal. For both outcomes, there was no significant difference found (DGE: OR 0.33, 95% CI: 0.08 to 1.30, I 2 = 0%, nasogastric tubes removal: MD 4.08 days, 95% CI: −10.74 to 2.58, I 2 = 99%). The certainty of evidence was very low for both comparisons.
Gastric motility was reported differently in the included trials. Erythromycin nonsignificantly reduced the proportion of retention of liquids at 30 min (p = 0.07) and was associated with a significant improvement in solid emptying (20-120 min, NA, p < 0.01) when tested with radionucleotide gastric-emptying trials reported by Yeo et al. Only

Proton Pump Inhibitor
Two trials were found that reported on PPI after pancreatic resections (Table 4) [7,9]. Both trials investigated patients after pylorus-preserving pancreatoduodenectomy and were judged with a high overall risk of bias. Pooling was not possible as the trials did not report on the same outcomes of interest.
The trial of Toyota et al. analyzed 24 patients that received 20 mg of dissolved Omeprazole through a jejunal feeding tube twice daily for 7 days after surgery [7]. The control group consisting of 18 patients received the same volume of saline as a placebo (six exclusions due to gastric bleeding). Baseline characteristics were equally distributed. Eight complications were reported in the PPI-Group whereas seventeen were reported in the control group. Notably, there were no gastric bleedings in the PPI group compared to six in the control group (p < 0.05). All other reported complications were equally distributed.
Jang et al. included 18 patients receiving 30 mg of lansoprazole daily for 12 weeks postoperatively [9]. A total of 19 patients served as comparisons in the control group where no additional treatment was reported. There was no difference in baseline characteristics and perioperative parameters.
During a modified Lundh meal test, gastrin serum concentrations were measured. Postoperative gastrin levels were approximately twice as high compared to preoperatively in the lansoprazole group (100 pg/mL vs. 50 pg/mL after 30 min, p < 0.001), whereas there was no difference in the control group. Pancreatic volume decreased significantly by 44.0% in the control group compared to 10.7% in the intervention group (95% CI: NA, p < 0.001). Nutritional status was reported by percentage of body weight lost after 3 months (4.5% vs. 9.9%, 95% CI: NA, p = 0.007), patients who regained 95% of their previous body weight (12 vs. 7 patients, p = 0.072) and mean triceps skin-fold thickness (−8.7% vs. −21.5%, 95% CI: NA, p = 0.047) in favor of the lansoprazole group.
Pancreatic endocrine insufficiency was defined as pancreoprivic diabetes or prediabetes (fasting plasma glucose 110-125 mg/dL and 2-h glucose level 140-199 mg/dL). This was the case in 3 out of 16 patients in the intervention group and 7 out of 15 in the control group. Additionally, serum insulin levels were higher in the lansoprazole group (21.1, SD 12.1 uU/mL, 101.0% of preoperative level) than in the control group (6.9 SD 1.6 uU/mL, 47.6% of preoperative level) 3 months postoperatively. Exocrine function showed a significantly lower level of stool elastase in the control group when compared to lansoprazole (24 (SD6) vs. 59 (12) ug/L, p = 0.009). Only pancreatic leakage was reported as a complication with no significant difference (4 (22.2%) vs. 3 (15.8%), p = 0.332).

Antidiabetic Drugs
Two trials were found that met the inclusion criteria (Table 5) [54,55]. Due to different treatments, pooling was not possible.  Patients with an artificial pancreas device (SGT-22; Nikkiso Company, Tokyo, Japan, n = 17) were compared to patients with a sliding scale method (n = 13) in the trial of Okabayashi et al. [54]. Thirty-day postoperative mortality was zero in both groups. Blood glucose levels were significantly different between 2 to 18 h after surgery comparing both groups with lower values in the artificial pancreas group compared to the sliding scale group (p < 0.05). Mean total insulin use was 107IU (SD 109) in the artificial pancreas group versus 8IU (SD 6) in the sliding scale group during the first 18 h after surgery (p < 0.001). Overall risk of bias was high due to some concerns in randomization, deviations from intended intervention, and outcome measurement.

Pancreatic Enzyme Replacement Therapy
There were eight trials reporting on PERT that met our inclusion criteria (Table 6) [56][57][58][59][60][61][62][63]. Study protocols were not comparable. Therefore, pooling was not possible. The intervention group was defined as PERT when compared to placebo or high dose PERT when compared to low dose. Risk of bias assessment resulted in low risk of bias in one trial, some concerns in five trials, and a high risk of bias in two trials.
Kim et al. included patients with a stool elastase level < 200 ug/g. The intervention group received 40,000 IU 3 times daily for 3 months beginning at the first outpatient follow up, usually 3-4 weeks after surgery (n = 118). The control group received a placebo capsule comprising lactose and cellulose instead (n = 119). Except for weight, the baseline characteristics were equally distributed. Compliance was 69.1% in the intention to treat analysis.
After 3 months of therapy or placebo, the difference in weight change did not reach statistical significance (−0.68 kg for the PERT group vs. −1.19 kg for the placebo group; p = 0.31) in the intention to treat analysis. Stool frequency was equal in both groups (1.  Neoptolemos included 35 patients for the final analysis of a cross-over study. High dose pancreatin was compared to standard dose pancreatin after a run-in period of 2 weeks and then for two 14 day periods. Results were taken at the end of each period. Results of both groups were summarized for final analysis (high dose 1 + high dose 2 and low dose 1 + low dose 2). Details can be seen in the Supplementary Materials.
Finally, the study group of van Hoozen et al. analyzed 11 patients after pancreatic resections (local resection-longitudinal pancreatojejunostomy) due to chronic pancreatitis. After a 4-week period where both groups received pancreatin, pancreatin therapy was either continued (n = 5) or replaced by placebo (n = 6) for another 4 weeks.
Comparing treatment outcomes at 8 weeks, patients randomized to receive placebo demonstrated significantly worse fat and total energy absorption than patients who continued to receive pancreatin supplementation (p < 0.02 and p < 0.02). No differences in stool frequency were seen in this trial.

Discussion
This systematic review and meta-analysis of perioperative medical management in pancreatic surgery shows less pancreatic fistulas when using somatostatin analogs and for glucocorticoids perioperatively. This result was, however, observed only with a moderate and very low certainty of evidence according to the GRADE Criteria. For other medical interventions, the evidence from randomized clinical trials is limited. The above analyzed medical interventions are discussed separately in the following passages. A regularly updated living meta-analysis can be found on www.emps.evidencemap.surgery (accessed on 3 January 2023) [2].

Somatostatin Analogues
Octreotide and other somatostatin analogues reduce the secretion of pancreatic enzymes through two different mechanisms [64]. The exocrine pancreas function is inhibited by the direct inhibition of pancreatic acinar cells as well as indirect inhibition through other pancreas-stimulating hormones such as gastrin. Less flow of digesting pancreatic enzymes through the pancreatic anastomosis after PD or less pressure at the distal end after DP through less pancreatic fluid production are thought to have a beneficial influence on healing [65].
Often-used medicaments in the included studies are octreotide and somatostatin. Somatostatin due to its short half-life time must be infused continuously whereas octreotide can be administered subcutaneously every eight hours [64]. Most of the included studies did administer somatostatin analogues postoperatively for seven days. In line with this regimen, the effect of somatostatin analogues decreases after seven days through adaption, desensitization, and tachyphylaxis [66].
In the present meta-analysis, somatostatin analogues did reduce POPF with a low certainty of evidence. This was shown in both PD and DP and may lead to a shorter hospital stay. This did, however, not influence mortality or other complications such as PPH, DGE, bile leak, or intraabdominal abscesses.
Limitations of this pooled analysis must be considered in the heterogeneity of treatment regimes. Octreotide was the most commonly used analogue but also somatostatin, pasireotide, and vapreotide were administered. Three studies administered an additional preoperative dose [31,39,44]. These studies, however, did not show a benefit of preoperative somatostatin administration when analyzed individually. One trial included six patients after TP [19]. Exclusion of this trial for POPF analysis did not relevantly affect the overall effects. Furthermore, definition of POPF has changed recently [14]. Pancreatic fistulas grade A, now defined as biochemical leaks, were not included in this analysis when reported in the included trials but were not always distinguishable as such. However, a sub-analysis on POPF Grade B and C according to the ISGPS definition confirmed a benefit of using somatostatin analogues [14]. Last, as shown in multiple studies, a surgeon's experience, and the treatment volume of the clinic have a relevant impact on surgical outcomes [67]. The heterogeneity of included patients in the analyzed trials may implicate different surgical volumes. The effect of additional treatment may be influenced by the a priori risk of development of POPF. The risks of POPF are higher with soft pancreatic texture and main pancreatic duct diameter ≤ 3 mm [68]. Common adverse effects include nausea, abdominal cramps, or diarrhea and can be treated adequately [69].
Therefore, the evidence for routine usage of somatostatin postoperatively is limited. Further research evaluating the effectiveness of somatostatin analogues in specific cohorts such as patients at high risk for POPF undergoing PD are needed. The new definition of the ISGPS creating risk classes should be used for such research [68].

Glucocorticoids
A possible cause for postoperative complications such as POPF and DGE after partial pancreatectomies may be postoperative pancreatic inflammation, mainly caused by pancreatic damage due to surgical manipulation at the transection line [52]. Animal models have shown that acinar cells are relevant in this inflammation cascade and that the severity of postoperative pancreatitis is reduced by perioperative glucocorticoids [70,71]. Both analyzed trials included only patients with high acinar-cell-rich transection lines, thus those groups of patients in whom the greatest effect is expected. The inflammation process at the transection line peaks within 4 h of resection and causes edema and activation of pancreatic enzymes compressing healing and duct obstruction [72]. Additionally, a retrospective analysis showed better survival rates when dexamethasone was administered perioperatively. However, there was no reporting on postoperative complications in that trial [73]. Complications often feared in the use of glucocorticoids are infectious complications and decreased anastomotic and wound healing. However, a meta-analysis on the use of perioperative glucocorticoids in oncologic abdominal surgery showed lower postoperative IL6 (POD1) and CRP (POD 3) values and additionally less postoperative infectious complications [74]. Anastomotic leakage was equal in both groups of the mentioned meta-analysis where, due to the publication date, the discussed pancreatic trials were not included. In accordance with the findings in the meta-analysis, there was no significant difference in infectious complications, wound healing, DGE, PPH, mortality, and LOS in both trials. Antila et al. showed a decreased rate in clinically relevant POPF (Grade B + C) in patients after distal pancreatectomy. The difference in POPF for patients after PD in the trial of Laainen et al. was insignificant. The evidence of these findings is however very low and limited due to the small sample sizes of the included studies. Further trials are needed to investigate the role of glucocorticoids in pancreatic surgery.

Erythromycin
Delayed gastric emptying after pylorus-preserving pancreatoduodenectomies is thought to have different mechanisms. Peritonitis from anastomotic leakages, ischemia of the pyloric muscle cells, damage to the vagal nerve, and reduced circulating motilin levels are discussed [75]. Furthermore, opioids that are often prescribed directly after pancreatic surgery for pain management increases severity of DGE [76]. In 2007, and therefore after the included studies were published, the ISGPS developed a standardized definition of DGE based on time to tolerance of solid food intake and nasogastric tube necessity [48]. Erythromycin, primarily invented as an antibiotic medicament, showed agonistic effects on the motilin receptor and therefore has a beneficial effect on resolving gastroparesis [77]. Erythromycin is associated with QT prolongation and interactions with other agents that are metabolized by CYP3A4 [77]. Due to different treatment protocols, pooling of the included studies was not possible. Erythromycin did reduce DGE in the trial of Ohwada et al. [53] and non-significantly in the trial of Yeo et al. [6]. However, certainty of evidence on earlier nasogastric tube removal and quicker progression to diet is low [53].
These results must be interpreted with caution as sample sizes were small and limited data exist overall for the efficacy of erythromycin in patients with gastroparesis. However, when DGE is present, erythromycin may be a valuable option.

Proton Pump Inhibitor
PPI inhibit hydrogen-potassium ATPase of gastric parietal cells and therefore lower gastric acid production [78]. This has shown to be prophylactic for gastric bleeding in the trial of Toyota et al. [7]. Not only acute bleeding but also the rate of marginal ulcers at the gastrojejunostomy can be reduced by antisecretory drugs such as H 2 -Receptor inhibitors and PPI [79]. Chronic irritation of the anastomosed jejunum due to the acidic environment, which it is worse at buffering than the duodenum due to lacking Brunner glands, could also lead to cancer formation at the anastomotic site [80]. For these reasons, PPI are routinely prescribed worldwide after PD [79].
An additional effect of PPI is drug-induced hypergastrinemia, a trophic stimulator of exocrine pancreatic cells [78]. In animal models, this has shown to be preventive for pancreatic atrophy [81,82]. These findings could also be shown in the trial of Jang et al. with less pancreatic atrophy and a better exocrine (stool elastase) and endocrine (insulin levels) function postoperatively [9]. The authors concluded that this had a decisive effect on a better nutritional status and less weight loss. Most results, however, failed to show a significant difference which may be due to the trials being underpowered.
Due to the prevention of anastomotic bleeding, postoperative administration of PPI after pancreatoduodenectomy should be considered.

Antidiabetic Drugs
Partial pancreatic resections pose a significant risk for endocrine pancreatic insufficiency as approximately one out of five patients will develop pancreatic endocrine insufficiency after PD or DP according to long-term evaluations [83]. With total pancreatic resection, insulin management is even more delicate due to the abundance of the peptide hormone insulin that is produced in the pancreatic beta-cells of the islets of Langerhans and the opposing hormone glucagon. This leads to significant morbidity and mortality [84]. Furthermore, hyperglycemia induced by surgical stress is thought to impair postoperative recovery [85]. The studies of Okabayashi and van Veldhuizen et al. showed better postoperative blood glucose control and no serious adverse events with closed loop systems [54,55]. However, long-term outcomes are lacking.
More research is needed for the development of new artificial pancreatic devices to avoid the long-term consequences of hyperglycemia and to impede the risk of hyperglycemia. Regardless to this, regular follow-ups for endocrine pancreatic insufficiency and consultation of an endocrinologist if endocrine pancreatic insufficiency is present is recommended.

Pancreatic Enzyme Replacement Therapy
After pancreatic resections, exocrine function may be impaired leading to steatorrhea and intestinal malabsorption [8]. Several trials did compare pancreatic enzymes to placebo or high dose versus low dose PERT. Intestinal fat absorption was significantly better in the PERT groups when compared to placebo [60,61]. Compliance did have a significant effect on weight loss [57]. To improve compliance, high dose PERT-capsules have been introduced to the market to lessen the capsules needed per day. However, there is no evidence on the benefit of high versus low dose pancreatin replacement as stated by Neoptolemus et al. [58].
These results indicate an advantage in intestinal absorption. Intestinal absorption may be even more important in patients with pancreatic surgery who might suffer from malnutrition due to pre-existing pancreatic exocrine insufficiency caused by chronic pancreatitis or who might have had significant weight loss due to oncological reasons [86]. Therefore, a patient-adjusted dose of PERT treatment should be initiated after pancreatic resections. The focus should be on improving compliance to therapy.

Conclusions
Many trials on treatment with somatostatin analogues were found and there is a potential reduction in POPF. However, further research evaluating the effectiveness of somatostatin analogues in specific cohorts such as patients at high risk for POPF undergoing PD are needed. Data on glucocorticoids and its role in reducing complications after pancreatic surgery are limited. Further studies are needed to confirm the potential benefit for POPF reduction. There is low evidence for the use of erythromycin and proton pump inhibitors. Trials on antidiabetic drug treatment to provide glucose control after pancreatic surgery show promising results but are limited to a few trials. PERT improves intestinal fat absorption and is recommended to avoid malassimilation because of exocrine pancreatic insufficiency.