Impact of enhanced recovery protocols after pancreatoduodenectomy: meta-analysis

Abstract

Background

This individual-patient data meta-analysis investigated the effects of enhanced recovery after surgery (ERAS) protocols compared with conventional care on postoperative outcomes in patients undergoing pancreatoduodenectomy.

Methods

The Cochrane Library, MEDLINE, Embase, Scopus, and Web of Science were searched systematically for articles reporting outcomes of ERAS after pancreatoduodenectomy published up to August 2020. Comparative studies were included. Main outcomes were postoperative functional recovery elements, postoperative morbidity, duration of hospital stay, and readmission.

Results

Individual-patient data were obtained from 17 of 31 eligible studies comprising 3108 patients. Time to liquid (mean difference (MD) −3.23 (95 per cent c.i. −4.62 to −1.85) days; P < 0.001) and solid (−3.84 (−5.09 to −2.60) days; P < 0.001) intake, time to passage of first stool (MD −1.38 (−1.82 to −0.94) days; P < 0.001) and time to removal of the nasogastric tube (3.03 (−4.87 to −1.18) days; P = 0.001) were reduced with ERAS. ERAS was associated with lower overall morbidity (risk difference (RD) −0.04, 95 per cent c.i. −0.08 to −0.01; P = 0.015), less delayed gastric emptying (RD −0.11, −0.22 to −0.01; P = 0.039) and a shorter duration of hospital stay (MD −2.33 (−2.98 to −1.69) days; P < 0.001) without a higher readmission rate.

Conclusion

ERAS improved postoperative outcome after pancreatoduodenectomy. Implementation should be encouraged.

Lay Summary

Enhanced recovery protocols consist of interdisciplinary interventions aimed at standardizing care and reducing the impact of surgical stress. They often include a short period of preoperative fasting during the night before surgery, early removal of lines and surgical drains, early food intake and mobilization out of bed on the day of surgery. This study gives a summary of reports assessing such care protocols in patients undergoing pancreatic head surgery, and assesses the impact of these protocols on functional recovery in an analysis of individual-patient data. The study revealed the true benefits of enhanced recovery protocols, including shorter time to food intake, earlier bowel activity, fewer complications after surgery, and a shorter hospital stay compared with conventional care.

Introduction

Enhanced recovery after surgery (ERAS) protocols consist of interdisciplinary multimodal perioperative interventions aimed at systematizing treatment, with orderly milestones reducing potentially deleterious consequences of perisurgical stress reactions and thereby enhancing the recovery process¹. Early studies in colorectal surgery have demonstrated shorter hospital stay and fewer postoperative complications after implementation of ERAS^2–6. The first recommendations on ERAS for pancreatoduodenectomy (PD) were published in 2012⁷ and updated in 2019⁸.

PD is a demanding procedure involving a complex resection and reconstruction with three or more surgical anastomoses. The associated surgical stress is extensive^9,10. Implementation of ERAS for PD has been associated with reduced hospital stay, morbidity, and rates of delayed gastric emptying (DGE). Forthwith, costs have been reduced without an effect on mortality or readmission rates across all age groups^11–20.

ERAS protocols have changed over time and guidelines have been re-evaluated regularly. Traditional measures such as duration of hospital stay do not adequately reflect functional recovery as defined today. Duration of stay is influenced by a variety of factors unique to individual countries and healthcare systems.

In this individual-patient data (IPD) meta-analysis the efficacy of ERAS on short-term outcomes was investigated in patients undergoing PD based on IPD from RCTs and observational studies.

Methods

Overview

This study was registered at PROSPERO (CRD42020197261). The systematic review is reported in accordance with the Preferred Reporting Items for a Review and Meta-analysis of Individual Participant Data statement²¹.

Search strategy and study selection

The Cochrane Library, MEDLINE, Embase, Scopus, and Web of Science were searched systematically for articles in English, French, and German published in any year up to August 2020. An experienced medical librarian assisted in the development of the search strategy. The detailed literature search strategy for each database is described in Appendix S1.

Two authors independently screened the abstracts and reviewed full-text articles of potentially eligible studies. Disagreements were resolved in discussion with the senior author. To optimize the literature search, the reference lists of studies identified primarily were browsed systematically for additional articles. Grey literature was not searched.

Eligibility criteria

RCTs and observational studies reporting the efficacy of ERAS in patients undergoing PD (classical (cPD) or pylorus-preserving (ppPD) PD) for any indication with at least 30 days of follow-up were included. Both cPD or ppPD were included as no definite scientific conclusion has been made regarding the superiority of either approach with regard to postoperative outcome²².

Studies reporting on four or fewer items as part of the ERAS or non-elective procedures were excluded. Non-comparative studies, those with fewer than 20 patients per arm, and studies of distal or total pancreatectomies were excluded. If two or more articles reported data from the same centre and interval, the publication with largest sample size or most recent cohort was retained.

Outcomes

Primary outcome measures were postoperative functional recovery elements (time to liquid intake, time to solid intake, time to bowel movement, time to removal of nasogastric tube, and time to removal of the last abdominal drain), overall postoperative morbidity, duration of hospital stay, and readmission. Secondary outcomes were major complications (Dindo–Clavien grade IIIa or higher), minor complications (Dindo–Clavien grade I or II), rate of postoperative pancreatic fistula (POPF), including subdivision into clinically relevant grade B or C fistulas, and DGE subdivided into grade B or C reinterventions and reoperations²³.

All outcomes were prespecified except for duration of follow-up. Minimum follow-up was set to 30 days but more extended follow-up periods of up to 90 days were also included.

Data collection and risk-of-bias assessment

Data extraction was undertaken using a case report form to facilitate the data collection process and to ensure consistency. Data were extracted independently by two investigators, and disagreements were resolved by consultation with the senior investigator. Extracted variables included study author, publication year, study interval, ERAS items, patients’ age, and compliance.

The corresponding authors of the studies included were invited to share anonymized IPD. Access to clinical trial data, codebooks, and clinical study reports was requested. Data sets were received in digital form. Inconsistencies were logged and resolved by querying the study authors. The newly created unified anonymized database was used for the analysis. To assess the risk of bias, the revised risk of bias in randomized trials (RoB 2) and the Risk Of Bias In Non-randomized Studies—of Interventions (ROBINS-I) tools were used^24,25.

Data analyses

R version 3.6.1 was used to undertake the analysis with the varameta and metafor packages²⁶. Point estimates calculated were risk and mean differences. Estimates of the sample mean and standard deviation from a median were calculated using the varameta package.

The threshold for statistical significance was set at P = 0.050. A two-stage approach was adopted for the IPD meta-analyses. IPD within studies generated summary measures, which were combined using standard meta-analytical methods²⁷.

The I² statistic was used to measure heterogeneity among trials in each analysis. I² was considered representative of the severity of heterogeneity and was interpreted using thresholds that have been recommended previously (0–40 per cent, likely minimal; 30–50 per cent, likely moderate; 50–90 per cent, likely substantial; and 75–100 per cent, likely considerable), along with the strength of evidence (P < 0.100 set for significance). A fixed-effect model was adopted if heterogeneity was not statistically significant and a random-effects model when statistical heterogeneity was identified²⁸. Missing values at the individual level were excluded.

To address IPD availability bias, aggregate data from trials in which IPD was sought but not obtained were added and pooled estimates calculated. Funnel plots stratified by IPD availability were created to visually explore the presence of publication bias and their symmetry was evaluated. In addition, Egger’s test was used to assess the presence of publication bias.

The robustness of the findings was examined in a priori-defined sensitivity analyses, in which the impact of exclusion of the most weighted study and lowest-quality studies was assessed.

A subgroup analysis was conducted for the two procedural variants (cPD and ppPD). A post hoc analysis of differences in postoperative bleeding and pulmonary complications between the ERAS and control groups and in the subgroups was carried out. A post hoc sensitivity analysis was also undertaken for studies published before and after 2015.

Results

Of 1160 studies identified in the literature search, 31 studies^{11–14,29–55} comprising 5844 patients were included (Fig. 1). All were single-centre studies. IPD was obtained from 17 of these studies with a total of 3143 patients. Data from 14 studies could not be obtained as they were no longer available in these centres. Of the studies with available IPD, two were randomized trials and 15 were non-randomized, retrospective analyses. Studies without IPD included three RCTs and 11 retrospective studies (Table 1). Twelve were from Europe, 12 from Asia, six from North America, and one from Australia. The results of risk-of-bias assessment are shown in Appendix S2. The ERAS protocols were heterogeneous and covered a time span of 14 years, with the majority specifying the following items: preoperative counselling, bowel preparation and fasting, antithrombotic and antimicrobial prophylaxis, postoperative nausea and vomiting prophylaxis, postoperative fluid infusions, analgesia, mobilization, oral intake, nasogastric tube, urinary catheter, abdominal drain removal, and discharge criteria (Appendix S3). The median number of items implemented was 16 (range 6–25). Compliance was reported in 10 of the 31 studies.

Fig. 1

PRISMA-IPD diagram showing selection of studies for review

PD, pancreatoduodenectomy; IPD, individual-patient data.

Table 1

Trial designs, patient characteristics, and compliance in included studies

Reference	Country	Study design	Study interval	No. of patients (intervention/control)	Compliance (intervention/control, %)	Disease entity
IPD available
Balzano et al.11	Italy	Observational	2004–2007 2000–2004	252/252	–	Cancer and benign
Braga et al.29	Italy	Observational	2010–2012 2008–2010	115/115	38–100 –	Cancer and benign
Coolsen et al.12	Netherlands	Observational	2009–2011 1995–2005	133/97	–	Cancer and benign
Daniel et al.31	USA	Observational	2015–2016 2013–2015	64/81	–	–
Hwang et al.33	Korea	RCT	2015–2017	123/124	–	Cancer and benign
Kagedan et al.34	Canada	Observational	2012–2014 2010	121/74	41–86 –	Cancer and benign
Lavu et al.39	USA	RCT	2015–2017	37/39	–	Cancer and benign
Lof et al.41	UK	Observational	2012–2017 2009–2012	250/125	26–99	Cancer and benign
Nussbaum et al.43	USA	Observational	2011–2012 2008–2010	100/142	–	Cancer and benign
Partelli et al.13	Italy	Observational	2013–2015 2009–2012	22/66	28–96 –	–
Passeri et al.44	USA	Observational	After 2015 Before 2015	86/86	–	Cancer
Pillai et al.45	India	Observational	2012 –	20/20	–	–
Su et al.48	China	Observational	2014–2015 2013–2014	31/31	39–100 –	Cancer and benign
Sutcliffe et al.49	UK	Observational	2012–2013 2011–2012	65/65	–	Cancer and benign
van der Kolk et al.51	Netherlands	Observational	2014–2016 2009–2012	95/52	> 80 –	Cancer and benign
Williamsson et al.53	Sweden	Observational	2011–2014	50/50	–	Cancer and benign
Zouros et al.55	Greece	Observational	2010–2014 2005–2009	75/50	73–100	Cancer and benign
Aggregate data only
Dai et al.30	China	Observational	2015–2016 2014–2015	68/98	–	Cancer and benign
Deng et al.32	China	RCT	2012–2014	76/83	90 95	Cancer
Joliat et al.14	Switzerland	Observational	2012–2014 2010–2012	74/87	68 82	Cancer and benign
Kapritsou et al.35	Greece	RCT	2012–2015	44/41	–	Cancer
Kuzmenko et al.38	Ukraine	Observational	2015–2017 2003–2014	39/39	–	Cancer and benign
Kim et al.36	Korea	Observational	2009–2010 2005–2008	88/185	–	Cancer and benign
Kobayashi et al.37	Japan	Observational	2010–2013 2005–2009	100/90	–	Cancer and benign
Li et al.40	China	Observational	2013–2017 2008–2013	203/141	–	Cancer
Nikfarjam et al.42	Australia	Observational	2009–2011 2005–2011	20/21	–	Cancer and benign
Shah et al.46	India	Observational	2008–2012 2006–2008	142/46	–	Cancer and benign
Shao et al.47	China	Observational	2011–2013 2009–2010	325/310	–	Cancer and benign
Takagi et al.50	Japan	RCT	2014–2016	37/37	43–100 –	Cancer and benign
Vanounou et al.52	USA	Observational	2004–2006 2001–2004	145/64	–	Cancer and benign
Zhu et al.54	China	Observational	2015–2017 2014–2015	64/69	–	Cancer

Individual-patient data (IPD) were obtained from 17 studies; aggregate data only were available for the other 14 studies.

Primary outcomes

Results for the five functional recovery elements assessed are shown in Fig. 2. Time to liquid and solid intake was shorter in the ERAS group, as was time to passage of first stool. Time to removal of the nasogastric tube was shorter in the ERAS group in all studies. The reported reinsertion rate from three studies ranged from 14 to 16 per cent in the ERAS groups and 5 to 26 per cent in the conventional group. The interval from surgery to removal of intra-abdominal drains was comparable in the two groups (Fig. 2 and Table 2). The overall complication rate was lower in the ERAS group (Fig. 3 and Table 2). Mortality was comparable between groups. Duration of hospital stay was shorter in the ERAS group, whereas readmission rates were similar in the two groups (Fig. 3 and Table 2). Publication bias assessment is shown in Appendix S4 and S5.

Fig. 2

Forest plots showing pooled estimated for functional recovery elements after pancreatoduodenectomy from six studies included in individual-patient data analysis

a Time to liquid intake, b time to solid intake, c time to passage of first stool, d time to removal of nasogastric tube (NGT), and e time to removal of the last abdominal drain. Effect estimates are shown with 95 per cent confidence intervals. ERAS, enhanced recovery after surgery; RE, random effects; FE, fixed effect.

Fig. 3

Forest plots showing estimates of postoperative complications and duration of hospital stay after pancreatoduodenectomy from individual-patient data analysis

a Overall complications, and b duration of hospital stay. Effect estimates are shown with 95 per cent confidence intervals. ERAS, enhanced recovery after surgery; FE, fixed effect.

Table 2

Pooled short-term outcomes of patients undergoing pancreatoduodenectomy

	Data analysed	Model	No. of studies	Heterogeneity		Pooled results
				I2 (%)	P	Effect estimate	P
Complications	IPD only	F	15	9.0	0.352	−0.04 (−0.08, −0.01)*	0.015
	IPD + AD	R	26	58.1	< 0.001	−0.10 (−0.14, −0.06)*	< 0.001
Mortality	IPD only	F	16	0	0.948	0 (−0.02, 0.01)	0.836
	IPD + AD	F	28	0	0.996	0 (−0.01, 0.001)	0.620
Time to liquid intake (days)	IPD only	R	4	83.7	< 0.001	−3.23 (−4.62, −1.85)*	< 0.001
	IPD + AD	R	8	94.7	< 0.001	−3.52 (−4.79, −2.25)*	< 0.001
Time to solid intake (days)	IPD only	F	3	27.6	0.251	−3.84 (−5.09, −2.60)*	< 0.001
	IPD + AD	R	7	85.7	< 0.001	−3.57 (−4.83, −2.31)*	< 0.001
Time to first stool passage (days)	IPD only	F	2	39.1	0.200	−1.38 (−1.82, −0.94)*	< 0.001
	IPD + AD	R	7	84.1	< 0.001	−1.32 (−184, −0.80)*	< 0.001
Time to NGT removal (days)	IPD only	R	4	87.9	< 0.001	−3.03 (−4.87, −1.18)*	0.001
	IPD + AD	R	6	91.8	< 0.001	−2.54 (−4.04, −1.03)*	< 0.001
Time to drain removal (days)	IPD only	R	6	53.5	0.047	−1.48 (−2.96, 0.01)	0.051
	IPD + AD	R	9	78.8	< 0.001	−2.63 (−4.13, −1.13)*	< 0.001
Duration of hospital stay (days)	IPD only	F	16	30.8	0.12	−2.33 (−2.98, −1.69)*	< 0.001
	IPD + AD	R	29	60.9	< 0.001	−3.57 (−4.21, −2.92)*	< 0.001
Readmission	IPD only	F	14	0	0.532	0.02 (−0.01, 0.05)	0.144
	IPD + AD	F	26	11.2	0.300	0.00 (−0.01, 0.01)	0.992
Major complications	IPD only	F	10	20.8	0.251	−0.01 (−0.05, 0.03)	0.708
	IPD + AD	R	14	43.6	0.031	−0.04 (−0.09. 0.001)	0.097
Minor complications	IPD only	F	10	11.4	0.251	0.01 (−0.03, 0.05)	0.708
	IPD + AD	R	14	27.9	0.009	−0.03 (−0.09, 0.02)	0.247
DGE	IPD only	R	14	93.5	< 0.001	−0.11 (−0.22, −0.01)*	0.039
	IPD + AD	R	25	90.1	< 0.001	−0.10 (−0.16, −0.04)*	0.001
DGE ≥ grade B	IPD only	R	10	92.1	< 0.001	−0.07 (−0.18, 0.04)	0.202
	IPD + AD	R	14	89.3	< 0.001	−0.06 (0.14, 0.02)	0.144
POPF	IPD only	F	16	0	0.878	−0.01 (−0.03, 0.02)	0.702
	IPD + AD	F	25	9.8	0.323	−0.02 (−0.04, 0.001)	0.082
POPF ≥ grade B	IPD only	F	13	0	0.984	−0.01 (−0.04, 0.01)	0.311
	IPD + AD	F	21	0	0.935	−0.02 (−0.04, −0.001)*	0.038
CT drainage	IPD only	R	6	48.0	0.069	0.01 (−0.05, 0.04)	0.730
	IPD + AD	R	6	48.1	0.069	−0.01 (−0.05, 0.04)	0.760
Relaparotomy	IPD only	F	15	0	0.492	0 (−0.02, 0.02)	0.757
	IPD + AD	F	19	0	0.562	0 (−0.02, 0.01)	0.552
Bleeding	IPD only	F	10	0	0.507	−0.01 (−0.03, 0.01)	0.529
	IPD + AD	F	17	0	0.691	0 (−0.01, 0.01)	0.557
Pulmonary complication	IPD only	F	8	29.9	0.189	−0.02 (−0.04, 0.01)	0.163
	IPD + AD	F	14	16.5	0.235	−0.01 (−0.03, 0.00)	0.123

Effect estimates (mean difference for continuous variables, risk difference otherwise) are shown with 95 per cent confidence intervals in parentheses. *Favours enhanced recovery protocol. Individual-patient data (IPD) were available for 17 of 31 studies. Also shown are the results of sensitivity analysis including aggregate data (AD) for the studies in which no IPD were obtained combined with the IPD. F, fixed effect; R, random effects; NGT, nasogastric tube; DGE, delayed gastric emptying; POPF, postoperative pancreatic fistula; CT, computer tomography.

Secondary outcomes

Major and minor complication rates were comparable in patients treated within an ERAS protocol and those who received conventional care. Development of POPF in general or clinically relevant POPF did not differ between the groups. DGE was less frequent in patients treated within an ERAS pathway, whereas grade B or C DGE was similar in the two groups. One study⁵¹ reported 68 per cent grade B and C DGE in the conventional group, which was much higher than in all other studies. Therefore, this study was omitted and the analysis rerun with a fixed-effect model, which did not show any difference between groups. Postoperative percutaneous drainage and re-exploration rates were no different between groups (Table 2).

A post hoc analysis of postoperative bleeding and pulmonary complications showed no difference between the groups.

Subgroup analysis by type of pancreatoduodenectomy

Time to recovery was consistently shorter in patients who underwent cPD within an ERAS protocol. There were insufficient data for a pooled analysis for ppPD (Appendix S5).

The overall complication rate was lower after ppPD treated within an ERAS protocol than with conventional care, whereas there was no difference among patients undergoing cPD with or without ERAS. For both cPD and ppPD, major and minor complication and mortality rates were comparable between the ERAS and conventional treatment groups. In both subgroups, postoperative hospital stay was shorter with ERAS, but readmission rates were comparable. ERAS was not associated with a different postoperative POPF rate after either type of PD. Development of DGE was less likely in patients who underwent ppPD within an ERAS protocol. Rates of grade B or C DGE were comparable in the two groups as were percutaneous drainage and reoperation rates and pulmonary complications were lower after cPD within ERAS.

Sensitivity analysis

The sensitivity analysis for morbidity and functional recovery showed no difference in outcomes except for time to drain removal and DGE grade B or higher, for which studies with aggregate data only favoured ERAS (Table 2). Regarding postoperative duration of hospital stay, in studies where IPD could not be obtained, the effect estimate was higher than for IPD studies. In a post hoc analysis, in which studies were split into early (before 2015) versus late published subgroups to assess the effect of improved ERAS protocols, results were similar to those of the primary analysis (Appendix S6).

Discussion

This meta-analysis showed a benefit of ERAS in time to functional recovery, overall postoperative morbidity, and postoperative length of hospital stay. In view of the advantages of ERAS, the implementation of such programmes should be encouraged in all centres performing PD.

Several other meta-analyses have focused on the effects of ERAS on postoperative outcomes. Hospital stay was reported to be shorter, and overall morbidity and DGE rates lower in the ERAS population, whereas rates of POPF, mortality, and readmission were comparable across the studies^15,17–19. Detailed investigation of DGE revealed no reduction in the rate of grade B or higher DGE in the ERAS populations¹⁷, which is in line with the results of the present analysis. The same applies to complications such as POPF, if graded individually according to severity, and to abdominal infections. Fewer pulmonary complications were observed in the subgroup of patients who underwent cPD within an ERAS protocol. This was concordant with the literature¹⁷. This effect was not, however, confirmed in the subgroup of patients undergoing ppPD.

Included studies revealed a benefit of ERAS for functional recovery elements, leading to the conclusion that functional recovery is faster if patients undergo PD within an ERAS pathway. Likewise, hospital stay was shorter in patients treated according to ERAS^18,19. Duration of hospital stay has gradually decreased over recent decades, influenced by local factors such as economic measures, cultural aspects, and improved patient workflow, which still vary greatly among countries. Therefore, functional recovery measures are a better indicator of the universal objective impact of ERAS and should be the primary outcome in RCTs, replacing length of stay^56,57. Complications and duration of hospital stay remain important to patients and stakeholders involved in healthcare. Hence, future studies should focus on prospectively capturing such information to address patient satisfaction, clinical results, and the cost-effectiveness of clinical care protocols.

Although ERAS improved the overall postoperative complication rate, stratification of complications according to severity (Dindo–Clavien grade I and II versus III and higher) did not reveal a benefit. This is in line with pooled results from a subset of studies^18,19. It is, however, difficult to draw a definitive conclusion with regard to the impact of ERAS on minor and major complications in this IPD meta-analysis because of the surprisingly lower major complication rate in studies in which IPD were not obtained. The analysis of major complications was, therefore, possibly underpowered.

The most consistent finding in the literature was a lower DGE rate after implementation of ERAS^15,18. Although some studies reported a higher DGE rate after a pylorus-preserving procedure, this IPD meta-analysis found an overall benefit for the subgroup of patients undergoing ppPD^58–60. As for complications according to severity, analyses of subgroups according to DGE grade were underpowered. Further comparative subgroup analyses on outcomes between cPD and ppPD showed little difference.

Limitations of this study include the design of the original studies, publication bias, and selection bias as IPD were not available from all studies. Aggregate data from all the included studies did not reveal a change in outcome results owing to limited IPD availability. Unfortunately, interstudy differences such as non-standardized outcome definitions or heterogeneous ERAS programmes could not be accounted for, nor could a uniform tool for measuring and reporting functional recovery after pancreatic surgery be proposed for future studies. As not all authors adhered to the proposed POPF and DGE definitions, the interpretation was somewhat complicated⁶¹. It was difficult to control for secular trends, which is a commonly occurring phenomenon in before-and-after studies, the design used for all the observational studies included in this review⁵². Finally, assessor blinding was not applied in any included study, including the randomized trials.

Funding

This study was funded by means of the Fondazione Poliambulanza Istituto Ospedaliero.

Acknowledgements

C.K. and C.T. are joint first authors of this review. This paper reports the results of a preregistered study, which can be accessed at https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42020197261.

Disclosure. The authors declare no conflict of interest.

Supplementary material

Supplementary material is available at BJS online.