Effect of prehabilitation on postoperative outcomes in patients with upper gastrointestinal tract cancer: meta-analysis

Abstract

Background

The aim of this meta-analysis was to elucidate the effects of prehabilitation (PR) on outcomes after surgery for upper gastrointestinal tract cancer.

Methods

PubMed, Web of Science, Embase, and Cochrane databases were searched from inception up to 21 May 2024 for randomized clinical trials (RCTs) and cohort studies investigating PR interventions in patients with upper gastrointestinal tract cancer. Data were synthesized using standardized mean differences (SMDs) and risk ratios (RRs) with corresponding 95% confidence intervals. Sensitivity and subgroup analyses were used to examine the robustness of the results and find possible sources of heterogeneity. Statistical analyses were performed using Review Manager 5.4 and Stata 16.0.

Results

Eight RCTs and eight cohort studies were included in the meta-analysis. Compared with the control group (no PR), the PR group had a significantly shorter postoperative length of hospital stay (SMD −0.27; 95% confidence interval (c.i.) −0.47 to −0.07; P = 0.008), a significant reduction in the occurrence of pneumonia after the surgery (RR 0.71; 95% c.i. 0.50 to 1.00; P = 0.005), and a greater improvement in the 6-minute walk distance (SMD 0.95; 95% c.i. 0.68 to 1.22; P < 0.00001). However, there were no significant differences between the control and PR groups in overall postoperative complications, anastomotic leakage, overall pulmonary complications, operative time, intraoperative blood loss, wound infection rate, in-hospital mortality, or recurrence rate (all P > 0.05).

Conclusion

For the population with upper gastrointestinal tract cancer, PR can partially lower the risk of postoperative pneumonia and promote faster postoperative recovery. Given the inherent limitations in the included studies, more large-scale RCTs are needed to verify these findings.

This systematic review and meta-analysis evaluated the clinical value of prehabilitation for patients undergoing surgery for upper gastrointestinal tumours. Eight randomized clinical trials and eight cohort studies were included. The results show that prehabilitation significantly reduced the postoperative length of hospital stay (standardized mean difference (SMD) −0.27), lowered the risk of pneumonia (risk ratio 0.71), and improved exercise tolerance (6-minute walk distance; SMD 0.95). However, it had no significant effect on other complications or indicators of long-term prognosis. This suggests that prehabilitation helps accelerate postoperative recovery, but further large-scale randomized clinical trials are needed to validate the clinical significance of the findings.

Introduction

Upper gastrointestinal tract cancer (UGIC) is a common malignancy that includes oesophageal and gastric cancers. UGIC is characterized by a high incidence rate and poor prognosis. According to China’s National Cancer Center, in 2022 there were approximately 224 000 new cases of oesophageal cancer and 358 700 new cases of gastric cancer in China, with these two malignancies ranking seventh and fifth, respectively, in terms of incidence^1,2. The corresponding number of deaths was 87 500 and 260,400, with mortality rates for oesophageal and gastric cancers ranked fourth and third, respectively, among all malignancies^1–3. Despite advances in oncological treatments, including radiotherapy and neoadjuvant chemotherapy, that may improve the prognosis of patients with UGIC, surgical resection remains the primary treatment⁴. However, the quality and speed of recovery after surgery for UGIC continue to be an issue in clinical practice.

Prehabilitation (PR) is a multidisciplinary preoperative strategy designed to optimize patients’ functional capacity before surgery. Focusing on the preoperative period, PR enhances patients’ cardiopulmonary functional reserves through optimized interventions⁵. A previous retrospective cohort study⁶on patients with oesophageal cancer indicated a significantly lower occurrence of postoperative pulmonary complications in the PR than non-PR cohort. In another study⁷, in which the effects of PR on postoperative outcomes after oesophageal cancer surgery were evaluated in propensity score-matched groups, PR significantly shortened the length of hospital stay (LOS) after oesophagectomy and reduced the incidence of pneumonia. A meta-analysis⁸ of the effects of perioperative rehabilitation, including PR and postoperative rehabilitation, on perioperative functional capacity, the incidence of postoperative pulmonary complications, postoperative morbidity, in-hospital mortality (IHM), LOS, and health-related quality of life in patients with oesophageal and gastric cancer found that preoperative interventions effectively decreased the occurrence and severity of postoperative pneumonia (PP; Clavien–Dindo grade ≥ II)⁸. These findings suggested potential benefits of perioperative rehabilitation for patients with oesophageal and gastric cancer. However, since the publication of that meta-analysis, inconsistent findings have been reported in seven new clinical studies^9–15. In addition, the generalizability of the findings of the meta-analysis⁸ to other UGICs, such as cancers of the oesophagogastric junction, and the efficacy of isolated PR interventions remain unknown.

Thus, a comprehensive synthesis of the most recent evidence regarding the role of PR in patients with UGIC is warranted. The aim of the present study was to clarify the effects of PR interventions on postoperative recovery outcomes in patients with UGIC via a systematic review and meta-analysis. The goal of the study is to present the latest and most thorough evidence-based medical insights to inform clinical practice.

Methods

Protocol and registration

This meta-analysis was registered with PROSPERO (Registration no. CRD42024574163). The results of this analysis are reported in accordance with the PRISMA guidelines and the PRISMA extension statement for meta-analyses.

Literature search

The PubMed, Cochrane, Embase, and Web of Science databases were searched for relevant articles using Medical Subject Headings (MeSH) and free-text terms (‘esophagus cancer’, ‘gastric cancer’, and ‘Prehabilitation’). The search strategy was formulated as (‘esophagus cancer’ OR ‘gastric cancer’) AND ‘Prehabilitation’. Studies meeting the inclusion criteria were screened. The search covered all records from inception up to 21 May 2024. Detailed search strategies are provided in the supplementary methods.

Eligibility criteria

Articles meeting the following criteria were eligible for inclusion:

• randomized clinical trials (RCTs) or cohort studies

• patients undergoing surgical treatment with pathologically confirmed oesophageal or gastric cancer

• an intervention group that received PR (including exercise and nutritional interventions, or exercise intervention alone—see below) in addition to standard care and a control group that received standard preoperative care

• studies reporting one or more of these following outcomes: total postoperative LOS, PP incidence, 6-minute walk distance (6MWD), overall complication rate, anastomotic leakage rate, incidence of pulmonary complications, operation time, intraoperative blood loss, wound infection rate, IHM, and recurrence rate.

The intervention methods for exercise included: those focused on functional abilities, such as walking endurance, strength training, and joint mobility; respiratory rehabilitation, including incentive spirometry, thoracic cage stretching, diaphragmatic breathing techniques, efficient coughing manoeuvres, and inspiratory muscle training; and aerobic conditioning through cycling and comprehensive muscle strengthening programs.

Articles were excluded if: they were cross-sectional studies, letters, conference abstracts, animal experiments, and review articles; they did not include PR as part of the intervention; they reported incomplete outcome measurements, with original data unavailable from the authors; and they were duplicates.

Data extraction

For each of the articles identified, two reviewers (J.L. and L.L.) independently gathered and carefully verified the author, year of publication, country, study population, sample size, mean age, sex distribution, PR interventions, and primary and secondary outcomes. Data for continuous variables were collected as the mean with standard deviation (s.d.). The number of events and total sample size was also collected. Disagreements were resolved by discussion with a third reviewer (Q.L.) to reach a consensus.

Quality assessment

The quality of eligible RCT studies was independently assessed by two reviewers (J.L. and L.L.), who then cross-verified their evaluations using Cochrane's Risk of Bias assessment tool. The assessment comprised random sequence generation, allocation concealment, participant and personnel blinding, completeness of outcome data, and reporting bias, among others. Three categories were used to examine methodological quality: high risk of bias, low risk of bias, and unclear bias. In the case of disagreement between the two reviewers, the final decision was made by a third reviewer (Q.L.). The quality of cohort studies was evaluated using the Newcastle–Ottawa Scale (NOS) in three domains: Selection, Comparability, and Outcome. The maximum possible score on the NOS is 9, with scores of 7–9 indicating high quality¹⁶.

Statistical analysis

References were managed using EndNote 9.0™ (Stamford, Connecticut, USA), data were organized using Excel® (Redmond, Washington, USA), and statistical analyses were performed using Review Manager 5.4™ (Oxford, UK) and Stata 16.0™ (College Station, Texas, USA). Effect sizes for categorical variables are reported as the risk ratio (RR), whereas those for continuous outcome variables are reported as the mean difference (MD) or standardized mean difference (SMD). Heterogeneity was assessed through 95% confidence intervals and I² statistics. Significant heterogeneity was indicated by P < 0.100 or I² ≥ 50%. All analyses were conducted via a random-effects model. In meta-analyses, the level of significance was set at α = 0.05. Sensitivity analyses were performed using the leave-one-out method. Funnel plots and Egger's test were used to determine publication bias. Subgroup analyses were conducted according to type of study, intervention duration, region, and type of cancer to examine the stability of the results and possible sources of heterogeneity.

Results

Literature search and results

The systematic retrieval and selection procedure is shown in Fig. 1. In all, 1158 pertinent articles were identified after thorough searches of PubMed (n = 326), Cochrane (n = 65), Embase (n = 178), and Web of Science (n = 589). After the removal of duplicates, 872 titles and abstracts were screened, with 842 studies being excluded, leaving 16 full-text articles for inclusion in the analysis^{6,7,9–14,17–23}. The 16 full-text articles involved 2085 patients (947 in the PR intervention cohort, 1138 in the control cohort) for pooled analysis. Eight of the studies^{9,10,13–15,21–23} were RCTs and eight^{6,7,11,12,17–20} were cohort studies. Seven studies^{6,12–15,17,23} were from Asia, four^7,9,10,18 were from Europe, two^20,21 were from the Americas, and three^11,19,22 were from Oceania. All studies were focused on the postoperative period in patients with gastric, oesophageal, or gastro-oesophageal junction cancer (Table 1).

Fig. 1.

Flow chart of the systematic literature search and selection process

RCT, randomized clinical trial.

Table 1.

Characteristics of the studies included in the meta-analysis

Author (year)	Study period	Region	Study population	No. of patients		Age* (years)		BMI* (kg/m2)		Sex (no. of male/female)
				T	C	T	C	T	C	T	C
Minnella et al.21 (2018)	February 2013–February 2017	Canada	Oesophagogastric cancer	26	25	67.3(7.4)	68.0(11.6)	26.1(4.8)	25.7(4.7)	18/8	20/5
Bausys et al.10 (2023)	February 2020–September 2022	UK	Gastric cancer	61	61	61(11)	64(10)	25.5(5.4)	27.1(4.9)	35/26	42/19
Min et al.14 (2024)	April 2022–February 2023	China	Oesophageal cancer	59	59	61.59(7.73)	60.95(8.53)			43/16	39/20
Yang et al.15 (2022)	June 2019–February 2020	China	Oesophageal cancer	107	108	63.09(8.98)	61.14(10.02)			72/35	81/27
Li et al.13 (2024)	January 2021–August 2022	China	Oesophageal cancer	68	68					45/23	48/20
Valkenet et al.22 (2018)	September 2013–May 2016	Netherlands	Oesophageal cancer	120	121	63.7(7.5)	62.7(8.9)	26.7(4.8)	26.5(5.2)	89/31	92/29
Allen et al.9 (2022)	December 2016–October 2018	UK	Oesophageal cancer	24	28	65(6)	62(9)	28.1(4.8)	27.7(5.1)	22/4	24/4
Yamana et al.23 (2015)	April 2011–December 2014	Japan	Oesophageal cancer	30	30	68.33(7.64)	65.90(9.50)	21.77(2.71)	20.91(2.53)	24/6	23/7
Kasai et al.12 (2024)	2016–2021	Japan	Oesophageal cancer	205	416	68.60(35.84)	68(41.66)	20.94(3.19)	21.18(3.07)	152/53	326/90
Janssen et al.11 (2022)	October 2015–February 2020	Netherlands	Oesophageal cancer	52	43	64(8)	65(9)	26.32(4.04)	25.73(4.76)	39/13	35/8
Akiyama et al.17 (2021)	September 2015–June 2019	Japan	Oesophageal cancer	23	25	65.9(7.7)	65.6(8.7)	21.7(2.5)	22.8(2.9) 0	17/6	21/4
Dettling et al.19 (2013)	January 2009–February 2010	Netherlands	Oesophageal cancer.	39	39	65.4(7.5)	66.5(9.6)	24.8(3.1)	25.9(2.9)	29/10	29/10
Halliday et al.7 (2021)	January 2016–April 2019	UK	Oesophageal cancer	38	38	67.22(10.02)	67.65(10.02)	27.5(5.6)	25.9(3.7)
Dewberry et al.20 (2019)	2014–2016	America	Oesophageal cancer	11	11	67.3 (mean)	62.7 (mean)			9/2	9/2
Inoue et al.6 (2013)	April 2008–November 2010	Japan	Oesophageal cancer	63	37	67.4(9.0)	65.0(7.8)	20.7(2.9)	21.1(2.7)	53/10	34/3
Christensen et al.18 (2019)	April 2016–May 2017	Denmark	Adenocarcinoma of the GEJ	21	29	63.9(8.2)	65.5(7.3)	28.4(5.6)	27.8(5.5)	18/3	27/2

*Data are presented as mean(standard deviation). BMI, body mass index; T, experimental group; C, control group; GEJ, gastro-oesophageal junction.

Risk of bias assessment

Of the eight RCTs, seven^{9,10,14,15,21–23} had a low risk of bias and one¹³ had a moderate risk of bias. Five studies^{10,14,15,21,22} reported computer-generated randomization, two^9,23 used sealed-envelope methods, and one¹³ did not describe the randomization method. Seven studies noted allocation concealment, whereas one study did not. Due to the nature of the interventions, blinding was challenging. Nevertheless, all studies implemented blinded outcome assessments. Three studies^13,14,23 did not specify whether selective reporting was present, which may introduce reporting bias. All outcome measures had complete data with clearly reported follow-up or dropout information (Figs 2, 3). The NOS scores for the eight cohort studies^{6,7,11,12,17–20} were in the range 7–9, reflecting high quality. The quality assessment results for the cohort studies are presented in Table S1.

Fig. 2.

Risk of bias of analysis across included randomized clinical trials

Fig. 3.

Representative summary table of the risk of bias of the included randomized clinical trials

Outcome measures

Length of hospital stay

Data from 13 studies^{7,9–15,17–19,21,22}, including 2025 patients (917 in the PR cohort, 1108 in the control cohort), were investigated. Pooled analysis revealed a shorter LOS in the PR cohort (SMD −0.27; 95% confidence interval (c.i.) −0.47 to −0.07; P = 0.008) and notable heterogeneity (I² = 75%, P < 0.00001; Fig. 4a). The funnel plot indicated no publication bias (Fig. 7a). Egger's test indicated that there was no significant publication bias (P = 0.579; Fig. 8a).

Fig. 4.

Forest plots showing impact of prehabilitation on hospital stay, postoperative pneumonia, change in 6MWD, and incidence of complications overall

a Hospital stay, b postoperative pneumonia, c change in 6MWD, and d incidence of complications overall. Values in parentheses are 95% confidence intervals. s.d., Standard deviation; SMD, standardized mean difference; 6MWD, 6-minute walk distance; d.f., degrees of freedom.

Fig. 7.

Funnel plots

a Hospital stay, b postoperative pneumonia complications, c changes in 6MWD, d the overall incidence of complications, e anastomotic leakage, f the overall incidence of pulmonary complications, g operation time, h blood loss, i wound infection rate, j in-hospital mortality, and k readmission rate. SMD, standardized mean difference; RR, relative risk; 6MWD, 6-minute walk distance.

Fig. 8.

Egger's publication bias plots

a Hospital stay, b postoperative pneumonia complications, c changes in 6MWD, d the overall incidence of complications, e anastomotic leakage, f the overall incidence of pulmonary complications, g operation time, h blood loss, i wound infection rate, j in-hospital mortality, and k readmission rate. 6MWD, 6-minute walk distance.

Incidence of PP

In all, 1488 patients from 10 studies^{7,9–12,14,17–19,22} were chosen for the analysis of PP incidence (637 in the PR cohort, 851 in the routine care cohort). The PR group had a lower incidence of PP (RR 0.71; 95% c.i. 0.50 to 1.00; P = 0.005) and statistically significant heterogeneity (I² = 59%, P = 0.009; Fig. 4b). The funnel plot indicated minor publication bias (Fig. 7b), but Egger's test indicated no statistically significant publication bias (P = 0.246; Fig. 8b).

Change in 6MWD

Three studies^13,17,21 involving 235 patients (117 in the PR cohort, 118 in the routine care cohort) reported on changes in the 6MWD. The postoperative improvement in the 6MWD was markedly greater in the PR than control cohort (SMD 0.95; 95% c.i. 0.68 to 1.22; P < 0.00001), and there was no evident heterogeneity (I² = 0%, P = 0.64; Fig. 4c). The funnel plot indicated minor publication bias (Fig. 7c), but Egger's test revealed no publication bias (P = 0.707; Fig. 8c).

Overall incidence of complications

The overall occurrence of complications was analysed in eight studies^{7,10–12,14,17,18,21} involving 1175 patients (481 in the PR cohort, 694 in the routine care cohort). Pooled analysis suggested comparable postoperative overall complication rates across groups (RR 0.85; 95% c.i. 0.65 to 1.11; P = 0.23), marked heterogeneity (I² = 78%, P < 0.00001; Fig. 4d). The funnel plot showed slight asymmetry (Fig. 7d), but publication bias was not detected after Egger's test (P = 0.178; Fig. 8d).

Anastomotic leakage

Anastomotic leakage data were reported in six studies^{10,12,14,18,19,22} including 1222 patients (501 in the PR cohort, 721 in the routine care cohort). No significant difference was observed across the two cohorts, with pooled analysis indicating similar postoperative anastomotic leakage rates (RR 1.03; 95% c.i. 0.70 to 1.53; P = 0.87). No heterogeneity was noted (I² = 0%, P = 0.42; Fig. 5a). The funnel plot indicated no publication bias (Fig. 7e). Egger's test also indicated no publication bias (P = 0.742; Fig. 8e).

Fig. 5.

Forest plots showing impact of prehabilitation on anastomotic leakage, overall incidence of pulmonary complications, operation time, and blood loss

a Anastomotic leakage, b overall incidence of pulmonary complications, c operation time, and d blood loss. Values in parentheses are 95% confidence intervals. IV, inverse variance; M-H, Mantel–Haenzsel; s.d., standard deviation; SMD, standardized mean difference; d.f., degrees of freedom.

Incidence of overall pulmonary complications

Five studies^{7,11,17,22,23} involving 507 patients (93 in the PR cohort, 112 in the routine care cohort) reported data on the occurrence of overall pulmonary complications (including intraoperative and/or postoperative complications). The pooled analysis indicated that PR did not significantly alter the incidence of overall pulmonary complications, with no evident differences across groups (RR 0.84; 95% c.i. 0.56 to 1.24; P = 0.37). Statistically significant heterogeneity was detected (I² = 67%, P = 0.02; Fig. 5b). The funnel plot did not show publication bias (Fig. 7f), nor did Egger's test (P = 0.072; Fig. 8f).

Duration of surgery

Data on the duration of surgery were derived from five studies^6,9,10,21,23 involving 385 patients (204 in the PR cohort, 181 in the routine care cohort). PR did not affect the duration of surgery, with no notable difference across cohorts (SMD −0.08; 95% c.i. −0.29 to 0.12; P = 0.42) and no evident heterogeneity (I² = 0%, P = 0.49; Fig. 5c). The funnel plot indicated no publication bias (Fig. 7g), nor Egger's test did not identify publication bias (P = 0.579; Fig. 8g).

Intraoperative blood loss

Three studies^6,9,23 involving 212 patients (117 in the PR cohort, 95 in the routine care cohort) reported data on intraoperative blood loss. Pooled analysis indicated that PR did not affect intraoperative blood loss, with no marked difference between the two cohorts (SMD 0.01; 95% c.i. −0.26 to 0.29; P = 0.92) and no heterogeneity (I² = 0%, P = 0.79; Fig. 5d). The funnel plot appeared symmetrical (Fig. 7h) and Egger's test did not indicate publication bias (P = 0.539; Fig. 8h).

Wound infection rate

Five studies^{10,11,14,17,24} including 615 patients (310 in the PR cohort, 305 in the routine care cohort) reported postoperative wound infection rates in the two groups. Pooled analysis revealed comparable wound infection rates across cohorts (RR 0.89; 95% c.i. 0.36 to 1.48; P = 0.38), with no evident heterogeneity (I² = 10%, P = 0.35; Fig. 6a). Although asymmetry was noted in the funnel plot (Fig. 7i), Egger's test (P = 0.172) indicated no publication bias (Fig. 8i).

Fig. 6.

Forest plots showing impact of prehabilitation on wound infection rate, in-hospital mortality, and readmission rate

a Wound infection rate, b in-hospital mortality, and c readmission rate. Values in parentheses are 95% confidence intervals. M-H, Mantel–Haenzsel; s.d., standard deviation; SMD, standardized mean difference; d.f., degrees of freedom.

In-hospital mortality

Six studies^11,18–22 on 537 patients (269 in the PR cohort, 268 in the routine care cohort) were reported on IHM. Pooled analysis indicated that PR did not alter IHM, with no significant difference observed across cohorts (RR 0.86; 95% c.i. 0.34 to 2.15; P = 0.75; Fig. 6b). No significant heterogeneity was observed (I² = 0%, P = 0.83). The funnel plot did not show any apparent asymmetry (Fig. 7j), and Egger's test revealed no noticeable publication bias (P = 0.083; Fig. 8j).

Readmission rate

Data from six studies^{7,10,11,20–22} involving 601 patients (304 in the PR cohort, 297 in the routine care cohort) were investigated. PR did not influence the readmission rate, with no marked difference between the two groups (RR 0.74; 95% c.i. 0.41 to 1.35; P = 0.33). Heterogeneity was not significant (I² = 28%, P = 0.22; Fig. 6c). The funnel plot showed slight asymmetry (Fig. 7k). Nevertheless, Egger's test indicated no publication bias (P = 0.963; Fig. 8k).

Sensitivity analysis

A one-way sensitivity analysis was performed to compare LOS (Fig. S1A), pneumonia (Fig. S1B), complications (Fig. S1C), anastomotic leakage (Fig. S1D), incidence of pulmonary complications (Fig. S1E), operative time (Fig. S1F), blood loss (Fig. S1G), wound infection rate (Fig. S1H), IHM (Fig. S1I), 6MWD (Fig. S1J), and recurrence rate (Fig. S1K). The effects of each study on the pooled SMD and RR were evaluated by sequentially removing one study at a time. The removal of any single study did not substantially alter the newly pooled SMD or RR. Therefore, the results were considered robust.

Subgroup analysis

Subgroup analyses were performed according to population characteristics, including study design, duration of intervention, region, and type of cancer. Statistical methods were used to compare subgroup data to detect differences across six outcome measures: LOS, complications, pneumonia, anastomotic leakage, readmission rate, and IHM. Variations in intervention duration and population characteristics were major factors influencing the outcomes of PR. The subgroup analysis results are presented in Tables 2, 3.

Table 2.

Anastomotic leakage, readmission, and hospital mortality rates according to subgroup analysis

	Anastomotic leakage				Readmission rate (%)				Hospital mortality (%)
	No. of studies	RR	P	I 2 (%)	No. of studies	RR	P	I 2 (%)	No. of studies	RR	P	I 2 (%)
Total	6	1.03 (0.70, 1.53)	0.87	0	6	0.74 (0.41, 1.35)	0.33	28	6	0.86 (0.34, 2.15)	0.75	0
Study design
RCT	3	0.88 (0.52, 1.49)	0.63	0	3	0.54 (0.30, 0.97)	0.04	0	2	1.28 (0.35, 4.62)	0.71	0
Cohort	3	1.28 (0.65, 2.54)	0.48	14	3	1.15 (0.42, 3.13)	0.79	45	4	0.57 (0.15, 2.11)	0.40	0
Follow-up period
≥ 3 weeks	1	1.00 (0.06, 15.61)	1.0	NA	2	0.34 (0.10, 1.23)	0.1	0	1	0.32 (0.01, 7.53)	0.48	NA
< 3 weeks	2	1.43 (0.62, 3.29)	0.41	40	1	0.60 (0.31, 1.18)	0.14	NA	2	0.95 (0.21, 4.33)	0.94	32
Region
Asia	2	0.78 (0.33, 1.87)	0.58	46
Europe	2	0.62 (0.11, 3.45)	0.59	0	2	0.97 (0.10, 9.75)	0.98	92	1	0.45 (0.02, 10.64)	0.62	NA
Oceania	2	1.43 (0.62, 3.29)	0.41	40	2	0.63 (0.35, 1.11)	0.11	0	3	1.02 (0.34, 3.02)	0.98	0
America					2	0.62 (0.17, 2.29)	0.47	0	2	0.63 (0.08, 4.75)	0.65	0
Population
Gastric cancer	1	1.00 (0.06, 15.61)	1.0	NA	1	0.29 (0.06, 1.32)	0.11	NA
Oesophageal cancer	4	1.06 (0.64, 1.77)	0.82	32	5	0.85 (0.46, 1.57)	0.59	26	5	0.91 (0.35, 2.38)	0.85	0
Adenocarcinoma of the GEJ	1	0.46 (0.05, 4.12)	0.49	NA					1	0.45 (0.02, 10.64)	0.62	NA

Values in parentheses are 95% confidence intervals. SMD, standardized mean difference; RR, risk ratio; RCT, randomized clinical trial; NA, not available; GEJ, gastro-oesophageal junction.

Table 3.

Length of hospital stay and the rate of complications and pneumonia according to subgroup analysis

	Length of hospital stay				Complications				Pneumonia
	No. of studies	SMD	P	I 2 (%)	No. of studies	RR	P	I 2 (%)	No. of studies	RR	P	I 2 (%)
Total	13	−0.27 (−0.47, −0.07)	0.008	75	8	0.85 (0.65, 1.11)	0.23	78	10	0.71 (0.50, 1.00)	0.05	59
Study design
RCT	7	−0.42 (−0.68, −0.17)	0.01	71	3	0.72 (0.27, 1.94)	0.52	86	4	0.83 (0.54, 1.27)	0.38	42
Cohort	6	−0.09 (−0.36, 0.18)	0.53	64	5	0.95 (0.78, 1.16)	0.64	60	6	0.66 (0.40, 1.11)	0.12	64
Follow-up period
≥ 3 weeks	4	−0.33 (−0.55, −0.11)	0.003	9	2	0.89 (0.18, 4.51)	0.89	93	2	0.52 (0.16, 1.69)	0.28	61
< 3 weeks	3	−0.04 (−0.53, 0.46)	0.89	76	1	0.79 (0.39, 1.61)	0.52		3	0.58 (0.23, 1.47)	0.25	81
Region
Asia	5	−0.54 (−0.89, −0.18)	0.003	85	3	0.81 (0.47, 1.41)	0.47	70	3	0.79 (0.32, 1.92)	0.6	24
Europe	4	−0.22 (−0.52, 0.08)	0.14	38	3	0.73 (0.40, 1.34)	0.31	91	4	0.60 (0.32, 1.11)	0.11	58
Oceania	3	0.02 (−0.35, 0.40)	0.90	69	1	0.74 (0.51, 1.07)	0.11	NA	3	0.75 (0.40, 1.41)	0.37	78
America	1	0.02 (−0.53, 0.57)	0.94	NA	1	2.08 (1.02, 4.25)	0.04	NA
Population
Gastric cancer	1	−0.25 (−0.60, 0.11)	0.18	NA	1	0.40 (0.24, 0.66)	0.0004	NA	1	0.25 (0.06, 1.13)	0.07	NA
Oesophageal cancer	14	−0.32 (−0.54, −0.09)	0.005	77	6	0.89 (0.66, 1.20)	0.46	71	8	0.71 (0.50, 1.00)	0.05	62
Adenocarcinoma of the GEJ	1	0.31 (−0.26, 0.87)	0.29	NA	1	1.14 (0.90, 1.44)	0.27	NA	1	1.84 (0.46, 7.37)	0.39	NA

Values in parentheses are 95% confidence intervals. SMD, standardized mean difference; RR, risk ratio; RCT, randomized clinical trial; NA, not available; GEJ, gastro-oesophageal junction.

Discussion

PR measures, supported by strong evidence from clinical studies, have significantly reduced postoperative complications, shortened the duration of hospital stays, and lowered the risks of readmission and death by improving perioperative stress responses in patients. Since its introduction in 1997, PR has expanded beyond colorectal surgery to other fields, such as obstetrics and gynaecology, urology, pelvic surgery, and orthopaedics. In the case of gastric resection, colorectal procedures, and hepatobiliary and pancreatic surgeries, evidence-based research^24–41 strongly supports the benefits of PR pathways, showing improvements in perioperative safety, patient satisfaction, shortened postoperative stays, and reduced complication rates. However, in UGIC, there have been relatively few clinical studies of the benefits of PR.

In this meta-analysis, 16 clinical studies published between 2008 and 2023 were included involving 2085 subjects^{6,7,9–14,17–23}. The meta-analysis focused on perioperative complications and postoperative functional recovery. The results indicated that PR measures notably lowered the occurrence of PP. However, there was no statistically significant difference between the PR and non-PR groups in the overall complication rate (such as anastomotic leakage, intra-abdominal bleeding, and vocal cord paralysis). Therefore, preoperative PR, which helps patients improve their heart and lung function before surgery, can effectively reduce the risk of PP. However, its impact on other complications around the time of surgery remains unclear. This meta-analysis showed that patients who received PR also showed significant improvements in the 6MWD compared with those who did not. Because the 6-minute walk test is a key measure of heart and lung fitness, the results suggest that preoperative PR may have a positive effect on overall cardiopulmonary function.

The meta-analysis of Tukanova et al.⁸ proved that perioperative or postoperative physical therapy could reduce the prevalence of pneumonia and overall complications in people undergoing surgery for oesophageal or gastric cancer. Tukanova et al.⁸ also demonstrated that perioperative or postoperative rehabilitation could shorten the LOS while improving symptoms of dyspnoea and physical function. This finding aligns with the conclusions of the present study. However, the present study specifically focused on the effectiveness of preoperative PR in UGIC, which helped reduce variations caused by differences in treatment approaches. This further confirms that PR can be a safe and effective option for preoperative rehabilitation in UGIC, with strong potential for clinical use.

The primary findings of this study showed some variability, indicating that the effects of the intervention may differ among specific subgroups. To investigate these differences and better understand the factors potential influencing the results, and their generalizability, subgroup analyses were conducted. One important observation was that the duration of the intervention influenced the effect of PR on the LOS. Interventions lasting ≥ 3 weeks were associated with better outcomes than those lasting < 3 weeks. This suggests that extending the duration of PR may help improve postoperative recovery. Therefore, it is recommended that the duration of PR for patients with UGIC be maintained at ≥ 3 weeks to achieve optimal benefits. In addition, regional subgroup analysis revealed that PR was more effective in Asian populations, with the effects being less pronounced for patients from other regions. This may be related to the number of studies included, because lifestyle, dietary patterns, and higher rates of Helicobacter pylori infection contribute to a higher incidence of UGIC in Asia, especially East Asia. In contrast, smaller sample sizes in regions like Europe and the Americas may introduce a small-sample-size effect, potentially leading to false-negative results⁴².

PR was also found to be more effective in patients with oesophageal cancer than in patients with gastric cancer or junctional cancer. This may be due to the common occurrence of cachexia and muscle wasting in oesophageal cancer, which heightens the demand for PR⁴³. Furthermore, the more aggressive nature of gastric and junctional cancers, along with their more complex prognostic factors, may have contributed to the observed differences. Moreover, the incidence of oesophageal cancer is higher than that of gastric and junctional cancers, and this, coupled with the relatively small sample size, may have led to potential false-negative results because of small-sample-size effects^44,45. However, because subgroup analyses have certain limitations, these results should be interpreted with care. More high-quality studies are needed to confirm and build on the findings of the present study.

The core of PR lies in its enhancement of patients’ cardiopulmonary reserve and physical fitness through preoperative interventions. The aim of PR is to alleviate postoperative pain and functional impairments while promoting faster recovery. Central to this approach is exercise training, particularly moderate-intensity aerobic activities such as walking, brisk walking, and swimming. These exercises improve cardiopulmonary function and muscle strength, help patients better withstand surgical stress, reduce inflammation and oxidative stress, and ultimately lower the risk of postoperative complications. In addition to physical training, nutrition, psychological support, and preoperative education and counselling further contribute to overall health, immune function, and treatment adherence. These elements facilitate wound healing, fasten recovery, and minimize postoperative complications^46,47.

This study has several limitations. First, the intervention methods varied slightly across studies: some involved only exercise training, whereas others included both exercise training and nutritional support. In addition, differences in intervention duration, exercise types, and exercise intensities may have influenced outcomes. Second, only articles published in English were included and unpublished literature was excluded. Third, blinding could not be implemented in the case of exercise interventions, and the included studies did not specify whether a double-blind design was used, which may result in implementation bias. Fifth, the number of studies was limited, and some outcome measures, such as LOS, surgery duration, and efficacy assessments, were subjective. Sixth, because research data were limited, subgroup analysis by surgical approach could not be performed. Therefore, further research is needed to determine whether the effects of PR differ among patients undergoing different open versus minimally invasive procedures. Seventh, this study spans a broad time frame (2015–2024). Advances in surgical techniques and PR practices over this period may have introduced variability, underscoring the need for more contemporaneous data to more accurately assess the effectiveness of presurgical stabilization. Finally, this study included only a few studies focused on gastro-oesophageal junction adenocarcinoma, which makes it harder to draw strong conclusions about how effective PR is for patients with gastric cancer and this specific type of cancer.

This study presents evidence supporting the effectiveness of PR exercises in decreasing the LOS and incidence of pneumonia, as well as improving exercise endurance (6MWD) in patients with UGIC. However, the specific impact of PR on health-related quality of life for these patients requires further investigation. To better understand and boost the clinical application of PR in cancer treatment, further high-quality, large-sample RCTs are needed. Future research should analyse the type, duration, and intensity of PR to determine the specific parameters that yield optimal outcomes. With more comprehensive and reliable data, it will become easier to accurately assess and promote PR as an effective approach to improving the health and recovery of patients with UGIC.

In conclusion, PR measures reduced the incidence of postoperative pulmonary complications in patients who underwent UGIG surgery and effectively shortened the postoperative LOS. Subgroup analyses showed that differences in the duration of the PR programs and patient characteristics (region, type of cancer) were key factors affecting the results. However, considering the potential heterogeneity and small sample size, large-scale prospective multicentre studies are needed to confirm the effects of PR on patients with UGIC and to identify factors potentially influencing outcomes. Future studies can standardize PR protocols and measure outcomes such as health-related quality of life.

Funding

The authors have no funding to declare.

Author contributions

Qi Li (Writing—original draft preparation, Writing—review & editing, Conceptualization, Formal analysis, Investigation, Supervision, Resources), Liqing Li (Writing—original draft, Formal analysis, investigation), Yeli Luo (Writing—original draft, Formal analysis, Investigation) and Jianhong Liu (Conceptualization, Methodology).

Disclosure

The authors declare no conflict of interest.

Supplementary material

Supplementary material is available at BJS Open  online.

Data availability

All data supporting the findings of this study are available within the paper and the accompanying Supplementary material.