Analgesic Modalities in Patients Undergoing Open Pancreatoduodenectomy—A Systematic Review and Meta-Analysis

Background: This systematic review explored the efficacy of different analgesic modalities and the impact on perioperative outcome in patients undergoing pancreatoduodenectomy. Methods: A systematic literature search was performed on PubMed, Embase, Web of Science, Scopus, and Cochrane Library Database using the PRISMA framework. The primary outcome was pain scores on postoperative day one (POD1) and postoperative day two (POD2). The secondary outcomes included length of hospital stay (LOS) and specific procedure-related complications. Results: Five randomized controlled trials and ten retrospective cohort studies were included in the systematic review. Studies compared epidural analgesia (EA), patient-controlled analgesia (PCA), continuous wound infiltration (CWI), continuous bilateral thoracic paravertebral infusion (CTPVI), intrathecal morphine (ITM), and sublingual sufentanil. The pain scores on POD1 (p < 0.001) and POD2 (p = 0.05) were higher in the PCA group compared with the EA group. Pain scores were comparable between EA and CWI plus PCA or CTPVI on POD1 and POD2. Pain scores were comparable between EA and ITM on POD1. The procedure-related complications and length of hospital stay were not significantly different according to the type of analgesia. Conclusions: EA provided lower pain scores compared with PCA on the first postoperative day after pancreatoduodenectomy; the length of hospital stay and procedure-related complications were similar between EA and PCA. CWI and CTPVI provided similar pain relief to EA.


Introduction
The postoperative pain after open pancreatoduodenectomy (PD) is significant owing to big incisions and extensive abdominal dissection associated in many cases with preoperative pain and opioid use prior to surgery [1,2]. Inadequate pain control affects the neuroendocrine stress response, increases complication rates, and causes delayed functional recovery and prolonged hospital stays [3][4][5]. Pain management is a key component of enhanced recovery after surgery (ERAS) protocol after PD [6], but optimal pain control continues to be a challenge.
On the contrary, intravenous PCA, one of the most common strategies for pain relief following pancreatoduodenectomy, offers effective pain relief, but with higher opioid consumption. Furthermore, it does not promote hypotension; consequently, the need for vasopressors and fluids is lower than EA [5,12,27,28]. However, their related side effects such as respiratory depression, ileus, nausea, vomiting, and long-term opioid dependence after surgery [2,5,29], as well as the evidence of a link between opioids and the growth of cancer cells [2,30], have raised some concerns about their use. Thus, some strategies of sparing opioids other than EA have become more common in major abdominal surgery.
Continuous infusion of local anesthetic par transabdominal wound catheter or wall catheter (transversus abdominis plane block) associated with PCA provides effective pain relief in major abdominal surgery, reduced opioid consumption, and the risk of complications such as hypotension and block failure compared with EA [31,32]. The bilateral paravertebral block provides better pain relief and reduced opioid consumption in the early postoperative period of abdominal surgery [33]. However, few data are available on pancreatic surgery [33][34][35][36][37].
Single-shot intrathecal morphine (ITM) is an effective strategy for postoperative analgesia following abdominal surgery, having a prolonged duration of action and higher potency [38]. ITM reduces pain at rest and on movement on the first postoperative day with low complication risk [39]. ITM, as a part of a multimodal pain management strategy in major hepatopancreatic biliary surgery, provides effective pain relief, requires less fluid therapy, and reduces the length of hospital stay compared with EA [40][41][42]. However, sparse data are available on pancreatoduodenectomy [40,43,44].
Sublingual sufentanil tablets (SSTs) as an alternative to EA seem to be an attractive analgesic option for pain management in major abdominal surgery and pancreatic surgery [45,46].
The present systematic review and meta-analysis aimed to summarize and compare the efficacy and safety of different analgesic modalities in patients undergoing pancreatoduodenectomy.

Material and Methods
This systematic review and meta-analysis were reported in accordance with the recommendations from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [47]. No prior registration of study protocol was undertaken for this review.

Eligibility Criteria
The inclusion criteria were randomized controlled trials (RCTs) or observational cohort studies published in English language that compare two or more analgesic techniques and reported at least one outcome of interest.
Studies that included minimally invasive cases (laparoscopic or robotic) or those that included open pancreatic resection (pancreatoduodenectomy and distal pancreatectomy) were excluded unless subgroup analysis of pancreatoduodenectomy was available. Case reports, conference abstracts, previously reported systematic reviews, studies with no full text available, and studies with fewer than ten patients were excluded too.

Information Sources and Search Strategy
We performed a systematic search of the electronic databases PubMed, Embase, Web of Science, Scopus, and Cochrane Library Database from 1 January 1995 through 1 October 2022. The following search terms were used ("pancreaticoduodenectomy" OR "duodenopancreatectomy OR "pancreas surgery" OR "pancreatic resection" OR "pancreatectomy" OR "Whipple operation/resection" OR "pancreatoduodenectomy" OR "pancreatoduodenal resection") AND ("analgesia" OR "pain control" OR "pain management" OR "postoperative pain" OR "opioid analgesia" OR "narcotic" OR" regional analgesia" OR "epidural analgesia" OR "neuraxial analgesia" OR "spinal block' OR 'intrathecal block " OR "patientcontrolled analgesia" OR "wound catheter" OR "TAP block" OR "paravertebral block"). The complete literature search is provided in Appendix S1 (Supplementary Information).

Selection Process
Two independent reviewers (L.L., O.C.) screened the titles and abstracts of all articles for relevance. Eligible full-text articles were subsequently reviewed by two other independent reviewers (R.T., O.B.). In case of any disagreement between the two reviewers, a third reviewer (L.M.) helped make the selection. Manual screening of the reference lists in relevant articles was conducted to identify any additional articles.

Data Collection Process and Items
Data were extracted by two independent reviewers directly from articles (L.L., R.T.). The following study characteristics were collected: country, study type, recruitment dates, centers, type of pancreatic resection, and analgesic modalities.

Primary and Secondary Outcomes Measurements
The primary outcome measures were pain scores on postoperative day 1 (POD1) and day 2 (POD2). Pain scores were rated on the numerical rating scale (NRS) from 0 to 10, where 0 indicated no pain and 10 correlated to the worst pain possible. In articles that used VAS, these were converted to the corresponding number on the NRS [48].
The secondary outcome measures included length of hospital day (LOS) and specific complications: postoperative pancreatic fistula, bile leakage, delayed gastric emptying (DGE), postoperative ileus, and gastrointestinal bleeding.

Study Quality Assessment
All studies included in the final analysis were assessed by two reviewers (M.S., D.C.L.) The risk of bias was evaluated using the Cochrane Risk-of-Bias tool 2.0 [49] for RCTs and the ROBINS-I tool [50] for non-randomized studies.

Statistical Analysis
The meta-analysis was performed using R environment for statistical computing and graphics (R Foundation for Statistical Computing, Vienna, Austria), version 4.1.2 [51], and the meta-R package [52].
For qualitative outcomes, the number of events and the total number of subjects in both intervention groups were extracted. For quantitative outcomes, the mean and SD were extracted. Data presented in the graphs of articles were extracted by Web plot digitizer [53]. Several equations were used prior to analysis for the estimation of mean and SD from several studies that reported the median, range, interquartile range, and sample size [54]. We calculated the random effects estimates for all outcomes because we presumed the presence of clinical variability between studies. To measure the heterogeneity variance for continuous outcomes, we used limited maximum likelihood and computed the mean difference as effect size. For qualitative outcomes, we used the Mantel-Haenszel estimator and computed odds ratios as effect size. For all estimators, a 95% confidence interval was calculated too. Heterogeneity was assessed using the I2 statistic; a threshold of 50% suggested moderate heterogeneity and 75% indicated substantial heterogeneity [55]. In the case of significant heterogeneity, sensitivity analyses were used using the leave-one-out method. A p-value of <0.05 was considered significant. The possible publication bias risk was assessed by the Egger test.

Results
The literature search identified 1903 studies. After duplicate removal and screening of titles and abstracts, 1775 studies were excluded because of irrelevance for this study. Details of the screening process are illustrated in Figure 1. Forty-three full-text articles were assessed for eligibility; of these, five RCTs [23,27,35,36,45] and ten retrospective cohort studies were included [10,[16][17][18]20,22,24,40,43,44]. The reason for the exclusion of full text is presented in Figure 1.
from several studies that reported the median, range, interquartile range, and sample size [54]. We calculated the random effects estimates for all outcomes because we presumed the presence of clinical variability between studies. To measure the heterogeneity variance for continuous outcomes, we used limited maximum likelihood and computed the mean difference as effect size. For qualitative outcomes, we used the Mantel-Haenszel estimator and computed odds ratios as effect size. For all estimators, a 95% confidence interval was calculated too. Heterogeneity was assessed using the I2 statistic; a threshold of 50% suggested moderate heterogeneity and 75% indicated substantial heterogeneity [55]. In the case of significant heterogeneity, sensitivity analyses were used using the leave-one-out method. A p-value of <0.05 was considered significant. The possible publication bias risk was assessed by the Egger test.

Risk of Bias Assessment
The results of the risk of bias assessment are presented in Tables 2 and 3. Concerning randomized controlled studies, four had some concerns regarding the overall risk of bias and one study was at high risk of bias. The most important domains affected by bias were the outcome measurement and the selection of the reported result. Within the cohort studies, three studies [16,18,20] were considered to have an overall high risk of bias and seven studies [10,17,22,24,40,43,44] had a moderate risk of bias. The most frequent observation was a moderate risk of bias due to confounding, measurement in outcomes, and selection of the reported result.

Overall Risk of Bias
Klotz et al. [27] 2020 Low risk Low risk Low risk Some concern Low risk Some concern Mungroop et al. [35] 2016 Low risk Low risk Low risk Some concern Low risk Some concern Marandola et al. [23] 2008 High risk High risk Low risk Some concern Some concern High risk Groen et al. [45] 2022 Low risk Some concern Low risk Some concern Low risk Some concern Hutchins et al. [36] 2018 Low risk Low risk Low risk Some concern Some concern Some concern Table 3. Risk of bias for cohort studies according to the Robins-I tool.

PCA versus EA
Pain score on POD1 was reported in five studies [16][17][18]23,24], including a total of 570 patients (PCA: n = 166, EA = 404), of which one study was an RCT [23] and the others were retrospective cohort studies [16][17][18]24]. The mean pain score was significantly higher in the PCA group compared with the EA group (MD 1.13, 95% CI: 0.52-1.74, p < 0.001) (Figure 2a). There was a moderate heterogeneity (50%; 95% CI: 0-81.7%), p = 0.091) and it was explored with a leave-one-out sensitivity analysis (Supplementary Figure S1). No matter which study was removed, the results were robust and remained statistically significant, pointing in the same direction. Only Marandola et al. [23] had a randomized controlled design, and their results were statistically significant. A sub-analysis of the other studies with a non-randomized design offered an MD of 1.05 (95% CI: 0.42-1.68; p = 0.001), which was statistically significant. The randomized study observed higher differences compared with non-randomized designs.
One retrospective cohort study [20] reported the median pain score in 42 patients who underwent pancreatoduodenectomy (PCA: 24, EA: 18). They observed no differences in pain score on POD1 (1.8 versus 1.2; p = 0.30), but a higher pain score in the PCA group versus the EA group, on POD2 (2.3 versus 1.3; p = 0.03).

Regional Techniques versus EA
One RCT [35]

Duration of Hospital Stay
Length of stay (LOS) was reported in 1918 patients included in seven studies. Five studies [10,16,17,24,27] comparing PCA to EA (1701 patients) and two studies [40,43] comparing ITM to EA (217 patients) were included in meta-analysis.
The length of stay was shorter in PCA compared with EA but not statistically significant (MD 0.84; 95% CI: −1.93 to 3.6, p = 0.55) (Figure 4a). A significant heterogeneity was found (I2 92.2%, 95% CI: 84.7-96%, p < 0.001). The heterogeneity was explored with a leave-one-out sensitivity analysis (Supplementary Figure S3). The study of Pratt [16] was an outlier. The data in this study were presented as the median and range, with the maximum value being very high. Based on these values, we computed the mean and standard deviation, and the values were influenced by this. When removing the study of Pratt [16] from the analysis, the length of stay was longer in the PCA group compared with the EA

Duration of Hospital Stay
Length of stay (LOS) was reported in 1918 patients included in seven studies. Five studies [10,16,17,24,27] comparing PCA to EA (1701 patients) and two studies [40,43] comparing ITM to EA (217 patients) were included in meta-analysis.
The length of stay was shorter in PCA compared with EA but not statistically significant (MD 0.84; 95% CI: −1.93 to 3.6, p = 0.55) (Figure 4a). A significant heterogeneity was found (I2 92.2%, 95% CI: 84.7-96%, p < 0.001). The heterogeneity was explored with a leave-one-out sensitivity analysis (Supplementary Figure S3). The study of Pratt [16] was an outlier. The data in this study were presented as the median and range, with the maximum value being very high. Based on these values, we computed the mean and standard deviation, and the values were influenced by this. When removing the study of Pratt [16] from the analysis, the length of stay was longer in the PCA group compared with the EA group (MD 0.50; 95% CI: −75-1.74, I2 = 60%.) Any other removal during the sensitivity analyses did not reduce heterogeneity substantially. group (MD 0.50; 95% CI: −75-1.74, I2 = 60%.) Any other removal during the sensitivity analyses did not reduce heterogeneity substantially.

Risk of Bias across Studies
The Egger test, for all analyses, had p-values above the threshold of significance.

Discussions
The present systematic review and meta-analysis of analgesic management in patients undergoing pancreatoduodenectomy demonstrated that intravenous opioid analgesia (PCA) provided higher pain scores on postoperative day 1 compared with epidural analgesia. However, when continuous wound infiltration was added to opioid systemic analgesia (PCA morphine) and compared to epidural analgesia, no significant differences in pain scores were demonstrated. Furthermore, there were no significant differences in pain scores when comparing intrathecal morphine followed by iv PCA to EA for the first postoperative day. Continuous bilateral paravertebral infusion (CTPVI) provided similar pain relief as epidural analgesia on POD1 and POD2. The surgical complications and length of hospital stay were not significantly different according to the type of analgesia. LOS was reduced when comparing ITM followed by iv opioid PCA to PCA opioid alone.
In major abdominal surgery, and more relevant in upper gastrointestinal and hepatopancreato-biliary surgery, where procedures can be disabling and most patients undergo oncologic resection, poorly controlled postoperative pain is associated with increased morbidity, delayed functional recovery, prolonged hospital stays, and reduced quality of life [3][4][5].
Thoracic epidural analgesia was considered for decades as a gold standard for pain relief following open major abdominal surgery, providing a reduction in rest pain, but the evidence regarding the impact on dynamic pain was not conclusive [12,25,28]. A meta-analysis of RCTs [28] comparing epidural analgesia to other alternative analgesic techniques (mainly IV opioids) in major abdominal surgery showed that epidural analgesia might be associated with superior rest pain control in the first 24-48 postoperative hours. On the contrary, a recent Cochrane review [12] reported, in abdominal surgery, a modest reduction in rest pain in the first 24 h after surgery with the use of EA versus opioid PCA. However, in terms of movement, epidural analgesia provides a greater reduction in pain scores compared with systemic opioids.
In our meta-analysis following pancreatoduodenectomy, when comparing intravenous opioid analgesia (PCA) to epidural analgesia (EA), we looked at the effectiveness and safety of analgesia for the first and second days after surgery. Accordingly, higher pain scores were reported in PCA on both days, but the clinical relevance of the superiority of EA analgesia was reduced. The largest mean differences in pain scores (MD 1.13, 95% CI: 0.52-1.74, p < 0.001) between PCA and EA were observed in POD1 and can be considered to have some clinical relevance. On POD2, the mean differences in pain scores between PCA and EA were lower (MD 0.56, 95% CI: 0-1.11, p = 0.05) and can be considered of limited clinical relevance.
Marandola et al. [23], in an RCT including 40 patients who underwent pancreatoduodenectomy, demonstrated better pain relief on the first postoperative day in epidural analgesia compared with intravenous opioid analgesia. Klots et al. [27], in a multicenter RCT including 248 patients following pancreatoduodenectomy, reported low pain scores on the second postoperative day in both groups PCA versus EA and, in consequence, the effectiveness of both procedures.
Our results are similar to those presented by Groen et al. [56] in their meta-analysis of epidural analgesia in patients following pancreatoduodenectomy. Patients receiving EA had slightly lower pain scores (MD −0.50; 95% CI: −0.8 to −0.21; p < 0.001) on days 0-3 compared with intravenous opioid analgesia. A significant difference in pain score was reported on POD1 (MD −1.08; 95% CI: −1.66 to −0.55; p < 0.001). In addition, another meta-analysis by Akter et al. [57] regarding analgesic modalities in pancreatic resection showed that EA compared with PCA in pancreatoduodenectomy provides a similar level of postoperative pain relief on POD2 (MD −0.29; 95% CI: −0.83 to −0.25; p = 0.04).
The general benefits of epidural infusion in pancreatic surgery, regarding better postoperative pain scores than intravenous analgesia, are observed in the first two postoperative days in most of the studies [1,9,10,15,20,23,24,27]; however, they are diminished further on.
Despite improved analgesia, the clinical benefits of epidural analgesia in terms of reduced postoperative morbidity (pulmonary, cardiovascular, thromboembolic, infectious, and surgical complications) and, in consequence, the length of hospital stay (LOS) showed conflicting evidence in different studies [1,[11][12][13]15,21,24,28,58]. Patients on EA require increased perioperative fluid administration and vasopressors, which could influence the outcome.
In our study, we did not find any association between PCA or EA and gastrointestinal complications after pancreatoduodenectomy (pancreatic fistula, bile leakage, delayed gastric emptying, ileus, and gastrointestinal bleeding). Similar results regarding pancreatic fistula, ileus, bile leak, and delayed gastric emptying were reported by other meta-analyses [56,57] in patients receiving PCA or EA following pancreatoduodenectomy.
Regarding LOS, our analysis did not find a significant difference between PCA and EA (MD 0.84, 95% CI: −1.93-3.6, p = 0.55), but LOS was slighter longer in PCA when the study of Pratt et al. [16] was omitted from the analysis. The present evidence regarding the impact of pain management on the length of stay in pancreaticoduodenectomy is derived predominantly from several retrospective studies [10,16,17,24] and only one RCT [27] with significant heterogeneity between studies (I2 92.2%, p < 0.001). The RCT of Klots et al. [27] showed a similar duration of hospital stay (15.7 ± 7.6 days) in PCA and EA groups, knowing that there were no differences in gastrointestinal or other complications (infectious, pulmonary, cardiovascular, and renal) between the groups. On the contrary, a retrospective study by Jajja et al. [10] reported a significant difference in LOS in PCA compared with EA (11.44 ± 12.3 versus 9.11 ± 6.8 days, p = 0.003) in 748 patients that underwent pancreatoduodenectomy. On a multivariate logistic regression analysis of the predictors for LOS, the authors [10] reported the use of PCA as being a significant predictor of LOS alongside ileus and delayed gastrointestinal emptying (p < 0.001). Our results were comparable to those reported by Akter et al. [57] in their meta-analysis when comparing PCA to EA in a subgroup of patients following pancreatoduodenectomy-no significant differences in LOS between PCA and EA (MD 0.09; 95% CI: −0.25 to 0.42; p = 0.38; I2 0%) were shown. However, a significantly shorter LOS after pancreatoduodenectomy, in EA compared with PCA (MD −2.69; 95% CI: −2.76 to −2.62; p < 0.001; I2 99%), was reported by Groen et al. [56] in their meta-analysis, based on only two retrospective studies with a very high heterogeneity between studies.
Very attractive analgesic opportunities in open hepato-biliary-pancreatic surgery are provided by transabdominal wall catheters (TAWCs) with continuous infusion of local anesthetic as a part of a multimodal analgesic regimen. TAWCs, compared with epidural analgesia, provide a similar level of pain relief but with fewer complications [31,32,34,37]. Continuous bilateral thoracic paravertebral infusion provided similar pain relief to epidural analgesia in major abdominal surgery, but with less side effects [59,60], and appeared to be a promising analgesic technique in terms of efficacy and safety [33].
In the present review, one RCT [35] compared EA to CWI plus PCA, and one RCT [36] compared EA to continuous bilateral thoracic paravertebral infusion following pancreatoduodenectomy. Thus, Mungroop et al. [35] compared EA to continuous wound infusion (CWI) plus PCA morphine (CWI-PCA) in a subgroup of 36 patients undergoing pancreatoduodenectomy. The authors reported a non-inferiority of CWI versus EA regarding pain scores, patient satisfaction, and complications. CWI-PCA was associated with a lower mean cumulative vasopressor and opioid consumption. Hutchins et al. [36] compared EA to continuous bilateral thoracic paravertebral infusion (CTPVI) in patients undergoing pancreatoduodenectomy, and they found a similar pain score on POD1 and POD2.
A non-inferiority of continuous wound infusion plus opioid PCA versus epidural analgesia in terms of pain relief, length of hospital stays, and opioid consumption was reported by Spoto et al. [61] in an RCT that included 80 patients who underwent pancreatic resection (pancreatectomy over 70%). However, further randomized large trials are needed for a definitive conclusion regarding the role of TAWCs and paravertebral block in pancreatoduodenectomy as a new strategy for effective pain relief in the perioperative period with reduced opioid consumption and improved functional recovery. The use of wound catheters as an alternative to epidural analgesia is recommended by the ERAS guidelines of perioperative care for pancreatoduodenectomy [6].
Regarding the efficacy and safety of ITM compared with EA in open hepato-pancreatobiliary surgery, a recent review [41] showed that the pain scores were lower or similar to EA in the first 24 h post-surgery and the length of hospital stay was reduced. In our review, one retrospective study [43] compared the analgesic efficacy of ITM as a part of multimodal pain management to EA in pancreatoduodenectomy and reported no difference in mean pain score on POD1, but significantly higher pain scores were reported on POD2 in the ITM group. However, no significant difference in postoperative pain score (over five days), opioid use, and length of hospital stay was reported between ITM and EA. When ITM followed by iv opioid PCA was compared to PCA alone after pancreatoduodenectomy, one study [44] reported a shorter length of hospital stays in the ITM group.
Sublingual sufentanil (SST) showed promising results as an analgesic option following pancreatoduodenectomy when compared with iv opioid PCA or patient-controlled epidural analgesia [45]. In this first randomized study [45] on a subgroup of 21 patients who underwent pancreatoduodenectomy, similar pain scores were reported on postoperative days 1 to 3 between SST and PCA or PCEA.
There are some limitations of the present review and meta-analysis. Most of the included studies were small, retrospective, non-randomized, single-center trials; although we included five RCTs in the narrative review, only two could be included in our metaanalysis. Regarding the risk of bias, the randomized controlled trials had some concerns of bias, and one trial had a high risk of bias. The observational studies included in our meta-analysis had a moderate risk of bias, except three studies that had a serious risk of bias. The postoperative pain scores and the specific gastrointestinal complications were not assessed in all studies; thereby, a meta-analysis was possible for opioid PCA versus epidural analgesia. The pain scores with movement were not reported. In our study, for several outcomes, we found the between-study heterogeneity to range from moderate to high, being statistically significant in the pain scores on the POD2 comparison and the length of stay comparison. We explored this heterogeneity with posthoc leave-one-out sensitivity analyses and, in the case of pain on POD1, the results were robust. Because of the limited data available regarding the opioid-sparing strategy analgesia (EA versus PCA plus continuous wound infiltration with local anesthetic) and the intrathecal opioid analgesia in pancreatoduodenectomy, the meta-analysis of primary or secondary outcomes was not possible and was limited to a narrative review.
For clinicians, the present review and meta-analysis summarized the available evidence regarding pain management following pancreatoduodenectomy. Most of the evidence is centered around the efficacy and safety of EA, PCA, and other opioid-sparing strategies such as continuous wound infusion or continuous bilateral continuous paravertebral. In addition, ITM, as a part of multimodal analgesia, has emerged as a promising alternative to epidural infusion in pancreatic surgery, offering similar analgesia with an enhanced safety profile.
More large randomized controlled trials that pay close attention to limit their biases are needed to evaluate the real advantages and disadvantages of these analgesic techniques in pancreatoduodenectomy, not only on postoperative pain, but on postoperative outcomes in the settings of ERAS pathways.

Conclusions
EA provided lower pain scores compared with PCA on the first postoperative day after pancreatoduodenectomy; the length of stay and procedure-related complications were similar between EA and PCA. Continuous wound infusion added to opioid PCA, and continuous bilateral thoracic paravertebral infusion provided similar pain relief compared with EA in the first two postoperative days after pancreatoduodenectomy.