Non-drug perioperative interventions to reduce postoperative pulmonary complications after abdominal surgery: systematic review and meta-analysis

Abstract

Objective

To evaluate the effectiveness of perioperative non-drug interventions in reducing postoperative pulmonary complications (PPCs) in adults undergoing abdominal surgery.

Design

Systematic review and meta-analysis.

Data sources

Ovid MEDLINE, Embase, and Web of Science from database inception to January 2025 and updated in January 2026, with no language restrictions.

Study selection

Randomised controlled trials assessing the effectiveness of perioperative non-drug interventions for the prevention of PPCs in adults undergoing elective abdominal surgery under general anaesthesia, with clearly defined PPCs.

Main outcome measures

The primary outcome was the proportion of patients developing PPCs. Secondary outcomes included the proportion of patients with PPC subtypes according to European Perioperative Clinical Outcome definitions (respiratory infection, respiratory failure, pleural effusion, atelectasis, or pneumothorax) and hospital length of stay.

Data extraction and synthesis

Two reviewers independently screened studies, extracted data, and assessed risk of bias with the Cochrane RoB 2.0 tool. Data were synthesised using meta-analyses and trial sequential analyses, with the evidence certainty assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach.

Results

255 trials including 55 260 participants were included, evaluating 10 types of interventions with 39 subtypes for PPC prevention. PPCs occurred in 6467 (11.7%) participants across all included trials. High certainty evidence showed that low fraction of inspired oxygen (FiO₂) significantly reduced PPCs (risk ratio 0.81, 95% confidence interval 0.71 to 0.92). Moderate certainty evidence showed benefit for four intervention types: lung protective ventilation (risk ratio 0.66, 0.57 to 0.76), physiotherapy (0.55, 0.46 to 0.65), analgesia (0.73, 0.64 to 0.84), and nutrition (0.74, 0.63 to 0.87), with individualised positive end expiratory pressure, composite lung protective ventilation, early mobilisation, and epidural analgesia also showing benefit at the subtype level. Trial sequential analysis confirmed sufficient cumulative evidence for all the above interventions except early mobilisation. By contrast, goal directed haemodynamic therapy, targeted blood pressure management, restrictive fluid therapy, and postoperative bi-level positive airway pressure showed no evidence of benefit, with moderate certainty.

Conclusions

This synthesis establishes an evidence hierarchy for PPC prevention in abdominal surgery. Low FiO₂ is the only intervention supported by high certainty evidence and should be prioritised in clinical practice. Other beneficial strategies include lung protective ventilation, physiotherapy, analgesic techniques, and nutrition interventions. Conversely, the role of goal directed haemodynamic therapy—despite its widespread use—warrants reconsideration for PPC prevention. These findings facilitate prioritisation of effective interventions and development of evidence based guidelines.

Study registration

PROSPERO CRD42025637449.

Introduction

Postoperative pulmonary complications (PPCs) rank among the most common and serious adverse outcomes following major surgery, particularly abdominal procedures.^{1 2} Reported PPC rates after abdominal surgery vary from approximately 12% to 65%,^{3 4 5 6} depending on diagnostic criteria, surgical type, and patient risk profiles. Patients who develop PPCs have substantial increases in hospital length of stay by three to 11 days,^{4 7 8} and costs are increased by 1.5-fold to twofold.^{4 8} Beyond the economic burden, PPCs account for approximately 50 000 deaths annually in the US alone.⁹ The pathophysiology involves multiple mechanisms. Abdominal surgery impairs diaphragmatic excursion and reduces functional residual capacity, promoting atelectasis and secretion retention through global disruptions in respiratory muscle coordination. Profound neuromuscular blockade is commonly used to optimise surgical exposure, but incomplete reversal can further impair early postoperative respiratory muscle function.¹⁰ These effects are exacerbated by general anaesthesia, which induces intraoperative atelectasis and depresses respiratory drive, and compounded by postoperative pain, opioid analgesia, and early immobility, which hinder deep breathing and effective cough.^{11 12}

PPCs are more common than cardiac complications and impose higher incremental costs,⁴ yet they receive less systematic attention. Evidence based guidelines remain limited.^{13 14} Despite this, multiple interventions, such as prehabilitation,¹⁵ lung protective ventilation strategies,^{5 16 17} and postoperative non-invasive respiratory support,^{18 19} are widely used in clinical practice. Compared with drug interventions, these non-drug strategies target key mechanisms driving PPCs (for example, impaired lung expansion, ineffective mucociliary clearance, and reduced mobility), and, unlike drug treatments, are easily integrated into contemporary enhanced recovery after surgery pathways.²⁰ A landmark systematic review published in The BMJ in 2020 synthesised 117 randomised trials across non-cardiac surgery but found no intervention supported by high certainty evidence for PPC prevention.²¹ Notably, substantial clinical and methodological heterogeneity across diverse surgical populations limited the applicability of these findings to specific procedures such as abdominal surgery. Furthermore, this evidence base is now outdated: since 2018, multiple high quality randomised controlled trials have evaluated interventions specifically in abdominal surgery—including preoperative education,¹⁵ goal directed haemodynamic therapy,^{22 23} and individualised ventilation strategies²⁴—but remain unsynthesised. Consequently, an up-to-date, abdominal surgery specific synthesis is needed to support consistent practice and guideline development.

We therefore conducted a comprehensive systematic review with meta-analysis and trial sequential analysis to assess the effectiveness of non-drug interventions in reducing PPCs in adult patients undergoing elective abdominal surgery under general anaesthesia. This surgery specific synthesis updates the evidence base to provide definitive effect estimates and identify interventions warranting prioritisation in clinical practice and guidelines. We used rigorous methods including reproducible search strategies, dual independent screening and data extraction, certainty of evidence assessment, and trial sequential analysis. By focusing exclusively on abdominal surgery, we overcome the heterogeneity limitations of previous broad reviews and provide actionable evidence for this high risk surgical population.

Methods

This systematic review and meta-analysis was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement (see eMethod 1).²⁵ The protocol was prospectively registered in PROSPERO (CRD42025637449) and has been published.²⁶

Search strategy

We searched Ovid MEDLINE, Embase, and Web of Science from inception to 15 January 2025, including ahead-of-print records, with no language restrictions. The search was updated on 20 January 2026 using identical search strategies (full strategies in eMethod 2). When we encountered reports in languages that none of the review team could interpret, professional translators were engaged; the translated material was subsequently checked by a bilingual colleague to ensure accuracy. Searches combined controlled vocabulary (Medical Subject Headings (MeSH) or Emtree) and key free text terms for perioperative care and PPCs based on European Perioperative Clinical Outcome (EPCO) definitions (eMethod 3).²⁷ Reference lists of eligible articles and previous reviews were manually screened to identify additional studies. Grey literature was not systematically searched owing to resource constraints.

Inclusion and exclusion criteria

We included all studies that met the following population, interventions, comparisons, outcomes, and study design questions.

Population

Adults (≥18 years) undergoing elective abdominal surgery under general anaesthesia, with or without adjunct regional techniques, were eligible. We excluded studies involving paediatric populations (<18 years), emergency procedures, non-abdominal surgeries, organ transplantation, or surgeries performed using regional anaesthesia as the sole technique.

Intervention

We included studies evaluating non-drug interventions delivered according to predefined protocols (with specified timing, duration, and delivery methods) administered preoperatively, intraoperatively, or postoperatively, with the intention of preventing PPCs. We excluded trials evaluating multimodal care pathways (for example, enhanced recovery after surgery) as standalone interventions because their composite nature precludes attributing effects to individual components and increases clinical heterogeneity. We also excluded studies that focused solely on surgical technique modifications or on drugs (for example, intravenous morphine versus hydromorphone).

For analgesia interventions, we included non-drug techniques, such as regional anaesthesia and procedural analgesic approaches. We included intravenous lidocaine infusion in this category, as it represents a perioperative analgesic technique administered via a standardised protocol, comparable in nature to regional anaesthesia techniques, rather than a comparison of different opioid or analgesic drugs.

Comparator

Eligible comparators were usual/standard care, no treatment, placebo, sham therapy, or a minimal intervention control consistent with routine practice. Trials comparing two active interventions, neither of which reflected routine practice or a minimal intervention, were excluded. Any ambiguities in comparator definitions were resolved by team consensus.

Outcomes

Eligible randomised controlled trials had to pre-specify PPCs as an outcome. PPCs were defined using author specified criteria, validated diagnostic tools (for example, the Melbourne Group Scale),²⁸ or consensus definitions such as EPCO.²⁷ PPCs were assessed during the primary postoperative period as defined by each trial. We excluded studies assessing complications only in the immediate postoperative period (post-anaesthesia care unit or first 24 hours). Studies reporting only respiratory muscle strength or lacking PPC explicit definitions were also excluded.

Study type

We included only randomised controlled trials published as full text, peer reviewed articles. We excluded abstracts, conference proceedings, non-randomised or observational designs, reviews, protocols, and studies without an accessible full text.

Primary and secondary outcomes

The primary outcome was the proportion of patients developing composite PPCs. When a composite was not explicitly reported, we derived it from EPCO defined components (respiratory infection, respiratory failure, pleural effusion, atelectasis, and pneumothorax), as reported in the original studies. When trials reported multiple PPC subtypes without information on patient overlap, we first attempted to contact study authors to obtain composite data. When composite data were unavailable, we used the highest individual subtype proportion per trial arm as a conservative estimate, acknowledging that this approach likely underestimates the true composite PPC occurrence.

Secondary outcomes were EPCO aligned PPC subtypes (respiratory infection, respiratory failure, pleural effusion, atelectasis, and pneumothorax) and hospital length of stay. Given the number of secondary outcomes, we present respiratory infection and atelectasis in the main text, with results for the remaining PPC subtypes provided in the web appendix.

Study selection, data extraction, and risk of bias assessment

All identified records were imported into Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia; www.covidence.org) for de-duplication and screening. The overall selection process is presented in the PRISMA flowchart (fig 1).

Fig 1.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart of study selection process. PPC=postoperative pulmonary complication

Two reviewers independently screened titles and abstracts, assessed full texts against pre-specified eligibility criteria, extracted study data, and evaluated risk of bias. Study data were extracted using a piloted standardised form and included study identifiers, population characteristics, intervention and comparator details (type, timing, duration, and pre-specified subcategories), outcome definitions, PPC events (composite and subtypes), length of stay, and mortality. Risk of bias was assessed using the Cochrane RoB 2.0 tool.²⁹

To ensure consistency, discrepancies at any stage were resolved through discussion or adjudication by a third reviewer. Study authors were contacted via email when necessary to clarify missing or ambiguous data.

Data synthesis and analysis

We systematically categorised all non-drug interventions identified in included trials and summarised the number of studies per intervention type. We did meta-analyses for intervention types evaluated in two or more trials. Where sufficient data were available (at least two trials), we also did subtype specific meta-analyses to provide more detailed evidence on individual intervention components. We initially fitted all random effects models by using the restricted maximum likelihood (REML) estimator. When REML failed to converge for several secondary outcomes with extremely sparse data, we used the DerSimonian-Laird method instead. To maintain methodological consistency and avoid selective estimator use, we uniformly applied the DerSimonian-Laird method to all secondary outcomes whenever a convergence problem arose. We calculated risk ratios with 95% confidence intervals for dichotomous outcomes, including composite PPCs and individual PPC subtypes defined by EPCO criteria. Trials with zero events in both arms were excluded from the quantitative synthesis as they cannot contribute to pooled relative effect estimates. For studies with zero events in one arm, we applied a continuity correction of 0.5. For continuous outcomes (for example, length of stay), we calculated mean differences with 95% confidence intervals. When studies reported continuous data as medians with interquartile ranges or ranges, we estimated means and standard deviations by using the methods described in section 6.5.2.9 of the Cochrane Handbook for Systematic Reviews of Interventions (version 6.5, 2024).^{30 31 32}

We assessed statistical heterogeneity by using the I² statistic, with values of ≥40% indicating important heterogeneity.³⁰ Results are presented in forest plots to visualise individual study effects and pooled estimates. For meta-analyses including 10 or more trials, we assessed potential publication bias by using funnel plots and Egger’s test. When quantitative synthesis was inappropriate owing to insufficient data or substantial clinical heterogeneity, we provided narrative summaries of study characteristics and findings. We used R statistical software (version 4.5.1) for all statistical analyses.

Certainty of evidence assessment

We evaluated the certainty of evidence for the primary outcome by using the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) framework.^{33 34} For each body of evidence examining the same specific intervention type, we independently assessed the overall certainty as high, moderate, low, or very low, on the basis of five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias.^{34 35} Two reviewers independently assessed the certainty of evidence, with disagreements resolved through discussion or third party adjudication. We documented the rationale for downgrading or upgrading evidence certainty for each domain. Detailed GRADE assessment criteria are provided in eMethod 4.

Trial sequential analysis

To control for the risks of type I and type II errors arising from sparse data and repeated significance testing in cumulative meta-analyses, we did trial sequential analysis for the primary outcome to calculate the required information size and to inform the assessment of imprecision within the GRADE framework.³⁶ We predefined a 20% relative risk reduction (risk ratio 0.80) as the clinically meaningful effect, with α=0.05, power=80% (β=0.20), and O’Brien-Fleming monitoring boundaries. The required information size represents the total sample size needed to reliably detect or exclude the predefined effect. We considered evidence to be conclusive when the cumulative Z curve crossed the monitoring boundaries before reaching the required information size or when the required information size was achieved without crossing boundaries. We used the RTSA package (version 0.2.2) in R statistical software for trial sequential analysis, with detailed method provided in eMethod 5.

Subgroup analyses and sensitivity analyses

To explore potential sources of heterogeneity, we did pre-specified subgroup analyses for the primary outcome stratified by surgical site (upper versus lower abdominal surgery) and surgical approach (laparoscopic versus open surgery). We did these analyses for all intervention types with sufficient data, including both main intervention types and their respective subtypes. We used tests for interaction (Cochran Q test) to assess whether treatment effects differed between subgroups, using random effects models. Interaction tests required at least two studies per subgroup. Studies including mixed surgical sites or approaches (for example, both upper and lower abdominal surgery or both laparoscopic and open surgery) were excluded from subgroup analyses. To assess the robustness of our findings, we did sensitivity analyses for the primary outcome by excluding trials at high risk of bias according to the Cochrane RoB 2.0 assessment.

Protocol amendments and deviations

No amendments were made to the registered protocol. One deviation occurred: mortality was pre-specified in the PROSPERO registration as a secondary outcome; however, it was reported in too few trials and with heterogeneous follow-up windows (ranging from in hospital to 180 days) to permit meaningful pooling, so mortality was not meta-analysed.

Patient and public involvement

This study evaluated existing data from randomised clinical trials. Patients and the public were not involved in the design, conduct, reporting, or dissemination plans of this systematic review owing to the limited resources and time.

Results

Description of included studies

In our initial and updated search, we identified 22 716 records from electronic databases and six additional records retrieved via hand searching. After de-duplication, 19 901 titles and abstracts were screened, and 674 full text articles were assessed. We excluded 419 articles (predominantly owing to non-eligible outcomes), with the final analysis including 255 trials (encompassing 55 260 participants) (fig 1).

Trials were predominantly conducted in Europe (44%) and Asia (31%), with most being single centre studies (79%) and more than half (54%) including ≥100 participants (table 1). Risk of bias assessment showed that 49.0% had low risk, 34.1% raised some concerns, and 16.9% were at high risk (eFigure 1). Full details of each trial are provided in eTables 1-3.

Table 1.

Summary of characteristics of included randomised controlled trials. Values are numbers (percentages)

Characteristic	Trials (n=255)
Geographical region:
Europe	112 (44)
Asia	80 (31)
America	33 (13)
Oceania	10 (4)
Africa	10 (4)
Multi-continent	10 (4)
Study centre setting:
Single centre	201 (79)
Multicentre	54 (21)
Sample size:
20- 99	118 (46)
100-999	126 (49)
≥1000	11 (4)
Surgical site (upper/lower abdomen):
Lower abdomen	85 (33)
Upper abdomen	86 (34)
Both upper and lower abdomen	84 (33)
Surgical approach:
Open surgery	84 (33)
Laparoscopy	65 (25)
Both laparoscopic and open	106 (42)
Intervention phase:
Preoperative	24 (9)
Intraoperative	80 (31)
Postoperative	74 (29)
Combination of phases	77 (30)

We identified 10 main types of interventions, with lung protective ventilation being the most frequently studied (n=49),^{5 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84} followed by physiotherapy (n=36),^{15 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119} nutrition (n=35),^{120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154} goal directed haemodynamic therapy (n=33),^{22 23 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185} analgesia (n=24),^{186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209} miscellaneous interventions (n=22),^{210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231} postoperative respiratory support (n=18),^{19 24 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247} drainage strategies (n=14),^{248 249 250 251 252 253 254 255 256 257 258 259 260 261} restrictive fluid therapy (n=13),^{262 263 264 265 266 267 268 269 270 271 272 273 274} and low fraction of inspired oxygen (FiO₂) (n=11).^{6 275 276 277 278 279 280 281 282 283 284} These intervention types were further subdivided into 39 specific subtypes, each involving distinct methods (fig 2). Funnel plots showed generally symmetrical distributions across most interventions, although potential publication bias was observed for individualised positive end expiratory pressure, composite lung protective ventilation, postoperative respiratory support, and immunonutrition (eTable 4; eFigure 2).

Fig 2.

Trial distribution by intervention type and subtype. An interactive version of this graphic and downloadable data are available at https://public.flourish.studio/visualisation/28102943

Primary outcome

PPCs occurred in 6467/55 260 (11.7%) participants across all included trials. Meta-analysis results stratified by intervention type and subtype are shown in figure 3 and eFigure 3.

Fig 3.

Forest plot of perioperative interventions for effectiveness in reducing risk of postoperative pulmonary complications (primary outcome). An interactive version of this graphic and downloadable data are available at https://public.flourish.studio/visualisation/28008910

Quality of evidence varied substantially across interventions (table 2). Only low FiO₂ achieved high certainty evidence for PPC prevention. Moderate certainty evidence supported benefit for four intervention types (lung protective ventilation, physiotherapy, analgesia, and nutrition) and four subtypes (individualised positive end expiratory pressure, composite lung protective ventilation, early mobilisation, and epidural analgesia). Conversely, moderate certainty evidence failed to show a benefit associated with goal directed haemodynamic therapy, targeted blood pressure management, restrictive fluid therapy, and postoperative respiratory support using bi-level positive airway pressure. The remaining interventions associated with benefit were supported by low certainty evidence.

Table 2.

Summary of findings and certainty of evidence for perioperative interventions in reducing risk of postoperative pulmonary complications

Intervention*	Anticipated absolute effects† (95% CI)		Relative effect—risk ratio (95% CI)	No of participants (No of RCTs)	Certainty of evidence (GRADE)‡
	Risk with comparison	Risk with intervention
Lung protective ventilation	278 per 1000	183 (158 to 211) per 1000	0.66 (0.57 to 0.76)	6363 (49)	⨁⨁⨁◯ Moderate
Individualised PEEP	230 per 1000	150 (117 to 191) per 1000	0.65 (0.51 to 0.83)	2754 (18)	⨁⨁⨁◯ Moderate
Composite lung protective ventilation	345 per 1000	210 (166 to 273) per 1000	0.61 (0.48 to 0.79)	2512 (16)	⨁⨁⨁◯ Moderate
Recruitment manoeuvres	220 per 1000	161 (123 to 211) per 1000	0.73 (0.56 to 0.96)	760 (10)	⨁⨁◯◯ Low
Fixed PEEP	303 per 1000	127 (79 to 209) per 1000	0.42 (0.26 to 0.69)	187 (3)	⨁⨁◯◯ Low
Tidal volume	289 per 1000	353 (226 to 550) per 1000	1.22 (0.78 to 1.90)	150 (2)	⨁◯◯◯ Very low
Physiotherapy	217 per 1000	119 (100 to 141) per 1000	0.55 (0.46 to 0.65)	4854 (36)	⨁⨁⨁◯ Moderate
Breathing exercise	296 per 1000	139 (107 to 184) per 1000	0.47 (0.36 to 0.62)	2346 (16)	⨁⨁◯◯ Low
Prehabilitation	144 per 1000	98 (75 to 127) per 1000	0.68 (0.52 to 0.88)	1301 (8)	⨁⨁◯◯ Low
Incentive spirometry	155 per 1000	113 (76 to 167) per 1000	0.73 (0.49 to 1.08)	407 (4)	⨁⨁◯◯ Low
Early mobilisation	185 per 1000	103 (57 to 183) per 1000	0.56 (0.31 to 0.99)	314 (3)	⨁⨁⨁◯ Moderate
Inspiratory muscle training	367 per 1000	125 (48 to 315) per 1000	0.34 (0.13 to 0.86)	61 (2)	⨁⨁◯◯ Low
Nutrition	112 per 1000	83 (71 to 97) per 1000	0.74 (0.63 to 0.87)	5220 (35)	⨁⨁⨁◯ Moderate
Immunonutrition	124 per 1000	102 (81 to 128) per 1000	0.82 (0.65 to 1.03)	2044 (14)	⨁◯◯◯ Very low
Postoperative early feeding	83 per 1000	67 (49 to 92) per 1000	0.81 (0.59 to 1.11)	1962 (12)	⨁⨁◯◯ Low
Nutrition supplementation	125 per 1000	64 (42 to 96) per 1000	0.51 (0.34 to 0.77)	1092 (7)	⨁⨁◯◯ Low
Preoperative oral carbohydrate loading	267 per 1000	117 (45 to 304) per 1000	0.44 (0.17 to 1.14)	122 (2)	⨁⨁◯◯ Low
GDHT	83 per 1000	76 (66 to 87) per 1000	0.91 (0.79 to 1.05)	8561 (33)	⨁⨁⨁◯ Moderate
CO or SV guided	75 per 1000	67 (55 to 81) per 1000	0.89 (0.73 to 1.08)	5097 (12)	⨁⨁◯◯ Low
SVV or PPV guided	125 per 1000	120 (62 to 228) per 1000	0.96 (0.50 to 1.83)	841 (7)	⨁◯◯◯ Very low
SVV or PPV, and CO or SV guided	129 per 1000	104 (71 to 153) per 1000	0.81 (0.55 to 1.19)	778 (5)	⨁⨁⨁◯ Moderate
Analgesia	159 per 1000	116 (102 to 134) per 1000	0.73 (0.64 to 0.84)	4613 (24)	⨁⨁⨁◯ Moderate
Epidural analgesia	172 per 1000	131 (112 to 155) per 1000	0.76 (0.65 to 0.90)	3155 (15)	⨁⨁⨁◯ Moderate
Nerve block	83 per 1000	62 (41 to 95) per 1000	0.75 (0.49 to 1.14)	1046 (7)	⨁⨁◯◯ Low
Miscellaneous:
Probiotic, prebiotics, or synbiotics	110 per 1000	64 (32 to 125) per 1000	0.58 (0.29 to 1.14)	427 (5)	⨁⨁◯◯ Low
Mechanical bowel preparation	38 per 1000	30 (13 to 67) per 1000	0.79 (0.35 to 1.79)	2910 (4)	⨁◯◯◯ Very low
Chewing gum	55 per 1000	54 (39 to 75) per 1000	0.97 (0.70 to 1.35)	2423 (4)	⨁◯◯◯ Very low
Targeted blood pressure	48 per 1000	45 (31 to 63) per 1000	0.94 (0.66 to 1.33)	2611 (2)	⨁⨁⨁◯ Moderate
Weight loss	18 per 1000	8 (1 to 56) per 1000	0.48 (0.07 to 3.21)	334 (2)	⨁◯◯◯ Very low
Systemic warming	81 per 1000	52 (17 to 156) per 1000	0.64 (0.21 to 1.92)	211 (2)	⨁◯◯◯ Very low
Postoperative respiratory support	104 per 1000	72 (53 to 95) per 1000	0.69 (0.51 to 0.91)	6497 (18)	⨁⨁◯◯ Low
CPAP	91 per 1000	84 (71 to 97) per 1000	0.92 (0.78 to 1.07)	5615 (6)	⨁⨁◯◯ Low
HFNC	263 per 1000	121 (68 to 213) per 1000	0.46 (0.26 to 0.81)	305 (4)	⨁⨁◯◯ Low
Bi-PAP	241 per 1000	157 (89 to 277) per 1000	0.65 (0.37 to 1.15)	232 (4)	⨁⨁⨁◯ Moderate
PSV during recovery period	85 per 1000	112 (32 to 388) per 1000	1.32 (0.38 to 4.57)	345 (4)	⨁◯◯◯ Very low
Drainage	106 per 1000	109 (72 to 167) per 1000	1.03 (0.68 to 1.57)	2092 (14)	⨁◯◯◯ Very low
No nasogastric or nasojejunal tube	100 per 1000	87 (60 to 129) per 1000	0.87 (0.60 to 1.28)	944 (6)	⨁◯◯◯ Very low
No drain	58 per 1000	112 (70 to 180) per 1000	1.93 (1.21 to 3.09)	994 (6)	⨁⨁◯◯ Low
Early drain removal	455 per 1000	177 (55 to 582) per 1000	0.39 (0.12 to 1.28)	154 (2)	⨁◯◯◯ Very low
Restrictive fluid therapy	54 per 1000	38 (25 to 57) per 1000	0.71 (0.47 to 1.06)	4140 (13)	⨁⨁⨁◯ Moderate
Low FiO2	239 per 1000	194 (170 to 220) per 1000	0.81 (0.71 to 0.92)	3650 (11)	⨁⨁⨁⨁ High

Bi-PAP=bi-level positive airway pressure; CI=confidence interval; CO=cardiac output; CPAP=continuous positive airway pressure; FiO₂=fraction of inspired oxygen; GDHT=goal directed haemodynamic therapy; GRADE=Grading of Recommendations, Assessment, Development and Evaluation; HFNC=high flow nasal cannula; PEEP=positive end expiratory pressure; PPV=pulse pressure variation; PSV=pressure support ventilation; RCT=randomised controlled trial; SV=stroke volume; SVV=stroke volume variation.

^*Evidence certainty (GRADE) assessment was not done for miscellaneous interventions and “other” subtype interventions owing to heterogeneous intervention definitions and limited clinical interpretability. These results are reported descriptively in figure 3 and accompanying text.

^†Risk in intervention group (and its 95% CI) is based on assumed risk in comparison group and relative effect of intervention (and its 95% CI).

^‡GRADE Working Group grades of evidence—(1) High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. (2) Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. (3) Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. (4) Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

Trial sequential analysis confirmed sufficient cumulative evidence for FiO₂, lung protective ventilation, physiotherapy, analgesia, and nutrition at the type level and for individualised positive end expiratory pressure, composite lung protective ventilation, breathing exercises, and epidural analgesia at the subtype level (eFigure 4).

Interventions with high to low certainty evidence showing benefit

This section details interventions that showed benefit in the meta-analysis. Table 3 provides a comprehensive synthesis of these findings, integrating GRADE certainty assessments and trial sequential analysis outcomes for all interventions showing benefit.

Table 3.

Summary of findings for perioperative interventions showing benefit in reducing postoperative pulmonary complications

Interventions	No of trials	No of participants	RR (95% CI)	P value	I2	GRADE certainty*	TSA result†	Sensitivity analysis‡
Lung protective ventilation	49	6363	0.66 (0.57 to 0.76)	<0.001	46.9	Moderate	Firm evidence	Robust
Individualised PEEP	18	2754	0.65 (0.51 to 0.83)	<0.001	47.7	Moderate	Firm evidence	Robust
Composite lung protective ventilation	16	2512	0.61 (0.48 to 0.79)	<0.001	61.6	Moderate	Firm evidence	Robust
Recruitment manoeuvres	10	760	0.73 (0.56 to 0.96)	0.03	0.0	Low	Inconclusive	Not robust
Fixed PEEP	3	187	0.42 (0.26 to 0.69)	<0.001	0.0	Low	Inconclusive	Robust
Physiotherapy	36	4854	0.55 (0.46 to 0.65)	<0.001	27.0	Moderate	Firm evidence	Robust
Breathing exercise	16	2346	0.47 (0.36 to 0.62)	<0.001	49.1	Low	Firm evidence	Robust
Prehabilitation	8	1301	0.68 (0.52 to 0.88)	0.004	0.0	Low	Inconclusive	Robust
Early mobilisation	3	314	0.56 (0.31 to 0.99)	0.05	0.0	Moderate	Inconclusive	Robust
Inspiratory muscle training	2	61	0.34 (0.13 to 0.86)	0.02	0.0	Low	Inconclusive	Robust
Nutrition	35	5220	0.74 (0.63 to 0.87)	<0.001	0.0	Moderate	Firm evidence	Robust
Nutrition supplementation§	7	1092	0.51 (0.34 to 0.77)	0.001	0.0	Low	Inconclusive	Not robust
Analgesia	24	4613	0.73 (0.64 to 0.84)	<0.001	0.0	Moderate	Firm evidence	Robust
Epidural analgesia	15	3155	0.76 (0.65 to 0.90)	<0.001	0.0	Moderate	Firm evidence	Robust
Postoperative respiratory support	18	6497	0.69 (0.51 to 0.91)	0.01	42.2	Low	Inconclusive	Robust
High flow nasal cannula	4	305	0.46 (0.26 to 0.81)	0.007	4.4	Low	Inconclusive	Robust
Low FiO2	11	3650	0.81 (0.71 to 0.92)	0.001	20.9	High	Firm evidence	Robust

CI=confidence interval; FiO₂=fraction of inspired oxygen; GRADE=Grading of Recommendations, Assessment, Development and Evaluation; PEEP=positive end expiratory pressure; RR=risk ratio; TSA=trial sequential analysis.

^*High=very confident in effect estimate; moderate=moderately confident; low=limited confidence.

^†TSA assessed cumulative evidence sufficiency for 20% relative risk reduction (α=0.05; power=80%). Firm evidence=Z curve crossed monitoring boundary or reached required information size; inconclusive=insufficient cumulative evidence.

^‡Results after exclusion of trials at high risk of bias. Robust=effect remained significant; not robust=effect lost significance.

^§Nutrition supplementation includes early postoperative supplemental enteral or parenteral nutrition, preoperative oral nutrition supplementation, and combined approaches.

Eleven trials with 3650 patients compared low versus high FiO₂ during the perioperative period, showing a 19% reduction in PPCs with low FiO₂ strategies (risk ratio 0.81, 95% confidence interval (CI) 0.71 to 0.92; I²=20.9%).^{6 275 276 277 278 279 280 281 282 283 284} The evidence achieved high GRADE certainty, with conclusive trial sequential analysis. The low FiO₂ strategies showed variability in oxygen concentration targets and application timing. Most trials used an FiO₂ of 30% as the low oxygen target,^{6 275 276 277 278 279 280} although some studies used 35% or 40%.^{281 282 283 284} Control groups predominantly received an FiO₂ of 80%,^{6 275 276 277 278 279 280 284} with a minority of studies using 50% or 60% intraoperatively.^{282 283} Implementation timing differed: most protocols implemented low FiO₂ throughout the intraoperative period (from intubation to extubation),^{6 277 281 282 284} whereas others extended this intervention through the early postoperative phase (typically maintaining low oxygen concentrations for the first two hours after surgery via face mask or reservoir devices).^{275 276 278 279 280 283}

Lung protective ventilation was the most extensively studied intervention, with 49 trials enrolling 6363 patients showing reduction in PPCs (risk ratio 0.66, 95% CI 0.57 to 0.76; I²=46.9%).^{5 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84} Trial sequential analysis confirmed firm evidence, and GRADE assessment indicated moderate certainty evidence overall. Trial sequential analysis showed that the cumulative Z curve crossed the trial sequential monitoring boundary and also reached the required information size.

Eighteen trials (2754 patients) compared individualised versus fixed positive end expiratory pressure strategies, showing a 35% reduction in PPC risk for individualised strategies (risk ratio 0.65, 95% CI 0.51 to 0.83; I²=47.7%) with moderate certainty evidence.^{37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54} Trial sequential analysis confirmed sufficient cumulative evidence. Individualisation strategies predominantly used either driving pressure guided titration (targeting minimum driving pressure) or dynamic compliance guided protocols (selecting positive end expiratory pressure with optimal compliance). Four studies used novel approaches—electrical impedance tomography guided titration to identify the intersection of alveolar collapse and overdistension and lung ultrasound guided adjustment to minimise lung ultrasound scores.^{42 46 48 49}

Sixteen trials with 2512 participants showed that composite lung protective ventilation strategies reduced PPCs by 39% (risk ratio 0.61, 95% CI 0.48 to 0.79; I²=61.6%), with firm evidence from trial sequential analysis and moderate certainty GRADE assessment.^{5 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69} These composite approaches predominantly used three strategy combinations: positive end expiratory pressure with low tidal volume and recruitment manoeuvres (triple combination, most common), positive end expiratory pressure with recruitment manoeuvres alone (dual combination), or positive end expiratory pressure with low tidal volume. Typical protocols involved tidal volumes of 6-8 mL/kg predicted body weight, positive end expiratory pressure levels of 5-12 cm H₂O, and recruitment manoeuvres performed at regular intervals (most commonly every 30 minutes or hourly) and at key time points (after induction, after discontinuation of mechanical ventilation, and before extubation). Some studies used higher positive end expiratory pressure levels (12 cm H₂O) combined with recruitment manoeuvres or implemented stepwise positive end expiratory pressure adjustment during recruitment manoeuvres.^{66 68}

Ten trials (760 patients) examining recruitment manoeuvres alone showed reduced PPCs (risk ratio 0.73, 95% CI 0.56 to 0.96; I²=0%).^{70 71 72 73 74 75 76 77 78 79} Recruitment manoeuvre protocols varied considerably, typically applying peak pressures of 30-40 cm H₂O for 20-40 seconds at key time points or regular intervals. Three small trials (187 patients) evaluating fixed positive end expiratory pressure strategies (8-10 cm H₂O throughout surgery) showed reduced PPCs (risk ratio 0.42, 0.26 to 0.69; I²=0%).^{80 81 82} For both recruitment manoeuvres alone and fixed positive end expiratory pressure, trial sequential analysis showed inconclusive evidence, and GRADE assessment yielded low certainty evidence, precluding definitive conclusions.

Physiotherapy interventions were evaluated in 36 trials enrolling 4854 patients, showing a 45% reduction in PPCs (risk ratio 0.55, 95% CI 0.46 to 0.65; I²=27.0%).^{15 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119} The GRADE assessment yielded moderate certainty evidence, with trial sequential analysis confirming sufficient cumulative evidence.

Sixteen trials evaluated breathing exercises, showing a 53% reduction in PPCs (risk ratio 0.47, 95% CI 0.36 to 0.62).^{15 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99} Trial sequential analysis confirmed firm evidence, although GRADE certainty was low, downgraded for risk of bias and inconsistency (I²=49.1%). Breathing exercise protocols varied in timing (preoperative only, postoperative only, or both) and modalities, encompassing lung expansion exercises, airway clearance techniques, and device assisted breathing. Training intensity ranged from hourly to multiple daily sessions, typically initiated two to seven days preoperatively when applied prophylactically and continued for several days postoperatively. Most interventions were supervised by physiotherapists or trained nurses, although some used standardised educational materials or video demonstrations.

Three trials examined early mobilisation, showing benefit in reducing PPCs (risk ratio 0.56, 95% CI 0.31 to 0.99; I²=0%).^{112 113 114} Evidence was graded as moderate certainty, but trial sequential analysis remained inconclusive. Protocols typically involved ambulation within 24 hours postoperatively with progressive increases in activity.

Eight trials evaluated prehabilitation programmes, showing reduced PPCs (risk ratio 0.68, 95% CI 0.52 to 0.88; I²=0%), although both GRADE assessment (low certainty) and trial sequential analysis (inconclusive) indicated limited evidence.^{100 101 102 103 104 105 106 107} Prehabilitation protocols consistently incorporated respiratory muscle training or breathing exercises across all trials, typically delivered over two to four weeks preoperatively. Most programmes used multimodal approaches combining respiratory training (inspiratory muscle training using threshold devices^{102 105 106} or breathing exercises^{100 101 103 104 106 107}), aerobic exercise (walking, cycling, or limb exercises),^{100 101 102 103 104 105 106} and resistance training (upper and lower extremity strengthening exercises).^{100 101 102 103 106 107} Delivery models typically involved regular supervised sessions and were supplemented with daily home based exercises. Some trials extended interventions into the postoperative period.^{104 105 106}

Two trials (61 patients) examining inspiratory muscle training reported a risk reduction (risk ratio 0.34, 95% CI 0.13 to 0.86; I²=0%), with a low certainty GRADE assessment and inconclusive trial sequential analysis.^{118 119} One trial implemented preoperative training in patients at high risk undergoing abdominal aortic aneurysm repair surgery,¹¹⁸ and the other applied postoperative training in patients undergoing hepatectomy.¹¹⁹

Analgesic techniques showed a 27% reduction in PPCs across 24 trials (4613 participants; risk ratio 0.73, 95% CI 0.64 to 0.84; I²=0%), with moderate certainty evidence and firm trial sequential analysis.^{186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209} Epidural analgesia was the most extensively studied subtype (15 trials; risk ratio 0.76, 95% CI 0.65 to 0.90; I²=0%), also rated as moderate certainty evidence.^{186 187 188 189 190 191 192 193 194 195 196 197 198 199 200} Most epidural trials compared thoracic epidural analgesia with systemic opioid based regimens, with consistent benefit observed across heterogeneous surgical populations.

Nutrition interventions across 35 trials (5220 patients) were associated with a 26% reduction in PPCs (risk ratio 0.74, 95% CI 0.63 to 0.87; I²=0%), supported by firm evidence from trial sequential analysis and moderate certainty evidence.^{120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154} Seven trials with 1092 patients found that perioperative nutrition supplementation reduced PPCs (risk ratio 0.51, 95% CI 0.34 to 0.77; I²=0%).^{120 121 122 123 124 125 126} However, GRADE assessment indicated low certainty evidence, and trial sequential analysis remained inconclusive. These interventions provided active nutrition support using standard enteral or parenteral formulations without immune enhancing components, delivered at various perioperative time points. The largest trial (468 malnourished patients with gastrointestinal cancer) showed that perioperative nutrition support (seven days preoperatively and postoperatively) reduced the risk of pneumonia.¹²⁶ Five trials implemented postoperative supplemental nutrition support using various delivery routes: enteral,^{120 121 123} combined enteral and parenteral,¹²⁴ and total parenteral.¹²⁵ One trial evaluated preoperative oral nutrition supplementation.¹²² These interventions provided active nutrition supplementation beyond routine care, distinct from interventions focused primarily on feeding timing (early feeding),^{127 128 129 130 131 132 133 134 135 136 137 138} specific nutrients (immunonutrition),^{139 140 141 142 143 144 145 146 147 148 149 150 151 152} and preoperative carbohydrate loading protocols.^{153 154}

Eighteen randomised controlled trials enrolling 6497 patients investigated postoperative respiratory support interventions, yielding an overall 31% reduction in PPCs (risk ratio 0.69, 95% CI 0.51 to 0.91; I²=42.2%), with low certainty evidence.^{19 24 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247} The cumulative Z curve did not cross the trial sequential monitoring boundary. Among the subtypes, only high flow nasal cannulas showed a significant benefit for PPC prevention.^{232 233 234 235} Other respiratory support modalities showed no significant benefits, with varying evidence certainty: bi-level positive airway pressure (moderate certainty evidence),^{236 237 238 239} continuous positive airway pressure (low certainty evidence),^{19 24 240 241 242 243} and pressure support ventilation during the recovery period (very low certainty evidence).^{244 245 246 247}

High flow nasal cannulas reduced PPC risk by 54% across four trials (305 patients; risk ratio 0.46, 95% CI 0.26 to 0.81; I²=4.4%), which was supported with low certainty evidence and insufficient evidence from trial sequential analysis.^{232 233 234 235} High flow nasal cannula protocols typically delivered heated and humidified oxygen at flow rates of 30-60 L/min with FiO₂ around 50%, initiated either immediately before extubation or on recovery unit admission, and continued for several hours on the basis of clinical response.

Interventions without evidence of benefit (moderate certainty)

Thirty three trials (8561 participants) compared goal directed haemodynamic therapy with usual care, showing no evidence of benefit for PPC reduction (risk ratio 0.91, 95% CI 0.79 to 1.05; I²=10.2%).^{22 23 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185} The evidence was graded as moderate certainty, and trial sequential analysis reached the required information size. Goal directed haemodynamic therapy protocols used diverse haemodynamic targets to guide perioperative fluid and vasopressor/inotrope administration, including cardiac output or stroke volume,^{22 23 155 156 157 158 159 160 161 162 163 164} stroke volume variation or pulse pressure variation,^{174 175 176 177 178 179 180} combined stroke volume variation/pulse pressure variation and cardiac output/stroke volume guided approach,^{181 182 183 184 185} and other targets such as inferior vena cava respiratory variability,¹⁶⁷ hypotension prediction index,¹⁶⁹ oxygen extraction ratio,¹⁶⁶ central venous oxygen saturation,¹⁶⁸ tissue oxygenation indices,¹⁷² and serum levels,¹⁷³ with monitoring achieved through various modalities such as pulse contour analysis (pulse index contour continuous cardiac output)¹⁷⁰ and oesophageal Doppler.^{165 171}

Among goal directed haemodynamic therapy subtypes, the combined stroke volume variation/pulse pressure variation and cardiac output/stroke volume guided approach showed no significant benefit with moderate certainty evidence (five trials, 778 participants; risk ratio 0.81, 95% CI 0.55 to 1.19; I²=0%).^{181 182 183 184 185} The five trials evaluating combined approaches used advanced haemodynamic monitoring integrating both dynamic (stroke volume variation/pulse pressure variation ≤12-13%) and static (cardiac index ≥2.5 L/min/m²) parameters and guided stepwise administration of crystalloid/colloid boluses, inotropes, and vasopressors according to predefined algorithms.

Two trials (2611 participants) evaluated intraoperative targeted blood pressure management strategies, using either individualised targets based on preoperative baseline or intensive management targeting a mean arterial pressure ≥80 mm Hg.^{230 231} Pooled analysis showed no evidence of benefit for PPC prevention (risk ratio 0.94, 95% CI 0.66 to 1.33; I²=0%; moderate certainty evidence).

Restrictive versus liberal perioperative fluid administration strategies were evaluated in 13 trials enrolling 4140 participants,^{262 263 264 265 266 267 268 269 270 271 272 273 274} yielding insufficient evidence of benefit for PPC reduction with restrictive fluid therapy on the basis of moderate certainty evidence (risk ratio 0.71, 95% CI 0.47 to 1.06; I²=21.9%). Restrictive regimens typically limited intraoperative crystalloid maintenance to 2-5 mL/kg/h with minimal or no preloading, compared with liberal groups receiving 7-18 mL/kg/h intraoperatively plus preloading volumes of 500-1000 mL.

Bi-level positive airway pressure protocols delivered both inspiratory (typically 12-16 cm H₂O) and expiratory (7-8 cm H₂O) positive pressure support, applied continuously for four hours in the recovery room, followed by intermittent use for one to three days postoperatively. All included trials enrolled predominantly obese or morbidly obese surgical populations.^{236 237 238 239} The pooled analysis of four trials showed no evidence of benefit for PPC prevention (risk ratio 0.65, 95% CI 0.37 to 1.15; I²=10.3%), with moderate certainty evidence.

Interventions without evidence of benefit (low to very low certainty)

Drainage strategies overall (14 trials; risk ratio 1.03, 95% CI 0.68 to 1.57; I²=55.5%)^{248 249 250 251 252 253 254 255 256 257 258 259 260 261} and miscellaneous interventions (22 trials; excluding targeted blood pressure management reported above; including probiotics, mechanical bowel preparation, chewing gum, weight loss, and systemic warming),^{210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229} both showed no significant effects with low to very low certainty evidence. However, substantial heterogeneity within drainage subtypes warranted separate evaluation (see below).

Interventions with potential harm

Complete omission of surgical drains was associated with higher PPC risk than drain placement (six trials, 994 participants; risk ratio 1.93, 95% CI 1.21 to 3.09; I²=0%).^{254 255 256 257 258 261} However, the low certainty of evidence—downgraded owing to risk of bias and imprecision—precludes definitive conclusions (eTable 4). Trial sequential analysis for this subtype showed insufficient information size, with the cumulative Z curve not crossing monitoring boundaries (eFigure 4).

Subgroup analysis

Subgroup analyses stratified by surgical site and approach showed generally consistent treatment effects across most interventions (eFigures 5 and 6). Although some interventions achieved statistical significance in one subgroup but not another, interaction tests were largely non-significant, indicating insufficient evidence that treatment effects differ between subgroups. The only statistically significant interaction was observed for stroke volume variation/pulse pressure variation guided goal directed haemodynamic therapy by surgical site (P=0.03), for which point estimates suggested opposite effect directions, although neither subgroup reached statistical significance individually.

Sensitivity analysis

Sensitivity analyses excluding trials at high risk of bias confirmed the robustness of the primary findings (eFigure 7). All six intervention types showing significant PPC reductions in the primary analysis—lung protective ventilation, physiotherapy, nutrition, analgesia, postoperative respiratory support, and low FiO₂—retained statistical significance with comparable effect sizes and directions. At the subtype level, most results were similarly consistent; notably, recruitment manoeuvres (risk ratio 0.78, 95% CI 0.52 to 1.15; five trials excluded) and nutrition supplementation (0.61, 0.34 to 1.09; three trials excluded) lost statistical significance. Conversely, the increased PPC risk associated with drain omission persisted and strengthened after exclusion of one high risk trial (risk ratio 2.22, 1.41 to 3.47).

Secondary outcomes

EPCO defined PPC subtypes

Meta-analysis of respiratory infection and atelectasis outcomes was based on fewer data (fig 4; fig 5) but showed variable patterns across interventions. Three intervention types showed significant benefits for both respiratory infection and atelectasis: physiotherapy, lung protective ventilation, and analgesia. However, most other interventions showed outcome specific effects. Postoperative respiratory support and low FiO₂ reduced atelectasis risk but showed no effect on respiratory infection. By contrast, goal directed haemodynamic therapy (risk ratio 0.80, 95% CI 0.66 to 0.97) and nutrition interventions (0.71, 0.59 to 0.86) reduced respiratory infection without showing significant benefit for atelectasis.

Fig 4.

Forest plot of perioperative interventions for effectiveness in reducing risk of respiratory infection. An interactive version of this graphic and downloadable data are available at https://public.flourish.studio/visualisation/28101870

Fig 5.

Forest plot of perioperative interventions for effectiveness in reducing risk of atelectasis. An interactive version of this graphic and downloadable data are available at https://public.flourish.studio/visualisation/28102172

Data on other PPC subtypes were more limited. Physiotherapy, analgesia, and composite lung protective ventilation significantly reduced respiratory failure (eFigure 8). For pleural effusion, only physiotherapy showed significant benefit, driven primarily by breathing exercises (eFigure 9). No interventions showed significant effects on pneumothorax, although events were rare across all trials (eFigure 10).

Hospital length of stay

When complete hospital length of stay data were available, trials were pooled and the results are presented in figure 6. Only nutrition (including its subtype of immunonutrition), cardiac output/stroke volume guided goal directed haemodynamic therapy, physiotherapy, and avoiding nasogastric or nasojejunal tubes showed reductions in hospital length of stay, although effect sizes were modest (ranging from 0.7 to 2.1 days). All other interventions showed no statistically significant differences in length of stay.

Fig 6.

Forest plot of hospital length of stay for perioperative interventions investigated to reduce postoperative pulmonary complications. An interactive version of this graphic and downloadable data are available at https://public.flourish.studio/visualisation/28102356

Narrative synthesis of single trial interventions

Five interventions were excluded from meta-analysis as each was evaluated in only a single trial. Full details are included in eTables 1-3. Three interventions were classified under the miscellaneous type, all reporting PPCs as secondary outcomes only. A trial investigated transcutaneous electrical acupoint stimulation started before anaesthesia induction and continued throughout surgery in patients undergoing gynaecological laparoscopic surgery, showing significantly reduced PPCs with improved oxygenation and reduced inflammatory markers.²²⁷ A cluster randomised trial evaluated preoperative patient education (a one hour seminar and brochure covering pneumonia prevention, mobilisation, and pain management) in patients undergoing major visceral surgery, showing no significant differences in pneumonia or pulmonary embolism between intervention and control groups.²¹⁷ Similarly, short term preoperative smoking cessation (counselling and nicotine replacement therapy for two to three weeks before surgery) showed no benefit on pneumonia in patients having colorectal surgery.²²⁴

Regarding the analgesic techniques, intrathecal opioids and intraoperative intravenous lidocaine infusion were each evaluated in single trials only. One trial compared intrathecal opioids (sufentanil with morphine) versus intravenous analgesia in patients undergoing abdominal aortic surgery, showing no differences in respiratory complications, including atelectasis, pneumonia, and respiratory failure.²⁰⁷ Conversely, intraoperative intravenous lidocaine infusion (a bolus of lidocaine 1.5 mg/kg before induction, followed by a continuous infusion of 2.0 mg/kg/h throughout surgery) in patients at intermediate or high risk for PPCs undergoing major abdominal surgery showed a significant reduction in composite PPCs (25.8% v 45.9%; risk ratio 0.56, 95% CI 0.38 to 0.84), particularly driven by lower atelectasis risk.²⁰⁸

Discussion

This systematic review of 255 randomised trials provides the most comprehensive surgery specific evidence synthesis for preventing PPCs in abdominal surgery, evaluating 10 intervention types encompassing 39 subtypes. Low FiO₂ emerged as the only intervention supported by high certainty evidence. Lung protective ventilation, physiotherapy, analgesic techniques, and nutrition at intervention type level, along with individualised positive end expiratory pressure, composite lung protective ventilation, early mobilisation, and epidural analgesia at the subtype level, showed benefit with moderate certainty evidence. With the exception of early mobilisation, all the aforementioned interventions were supported by firm evidence from trial sequential analysis. In particular, moderate certainty evidence suggested a lack of benefit for goal directed haemodynamic therapy, targeted blood pressure management, restrictive fluid therapy, and bi-level positive airway pressure. Notably, complete surgical drain omission was associated with a potential increase in PPC risk, although the low certainty evidence precludes definitive conclusions.

Clinical significance of PPC prevention

The 11.7% incidence observed across our included trials is consistent with previous meta-analyses of randomised controlled trials in this field (14.4%),²¹ translating to a substantial global burden given the estimated 313 million annual surgical procedures worldwide.²⁸⁵ The prolonged hospital admission, higher costs, and increased mortality associated with PPCs underscore the urgency of implementing evidence based preventive strategies.^{4 7 8 9} Our systematic synthesis establishes an evidence hierarchy that differentiates interventions by certainty of benefit, enabling prioritised clinical decision making and resource allocation.

Evidence supported strategies: high-to-moderate certainty evidence of benefit

Low FiO₂

Our analysis found high certainty evidence supporting lower FiO₂ in reducing PPCs, primarily driven by decreased atelectasis. Recent meta-analyses confirmed that low FiO₂ reduced atelectasis without affecting mortality.^{286 287 288} However, evidence on extrapulmonary complications remains conflicting: Wu and colleagues found no impact on cardiac or neurological complications,²⁸⁷ whereas Li and colleagues reported increased acute kidney injury with low FiO₂.²⁸⁶ These findings need cautious interpretation given the delicate balance between maintaining adequate tissue oxygenation and avoiding hyperoxia related complications. Implementation decisions should weigh the demonstrated PPC risk reduction against these potential extrapulmonary risks in individual patients. A registered large scale randomised controlled trial (ChiCTR2400083713) investigating titrated versus fixed high FiO₂ for PPCs in abdominal surgery may help to establish optimal perioperative oxygen concentration thresholds.

Lung protective ventilation

Both individualised positive end expiratory pressure and composite lung protective ventilation strategies are supported by moderate certainty evidence. Regarding individualised positive end expiratory pressure, although the recently published large scale DESIGNATION trial showed no significant effect,⁵¹ the pooled estimate remained protective against PPCs, consistent with previous meta-analyses,^{289 290} and trial sequential analysis confirmed the robustness of cumulative evidence. The principle of individualisation—tailoring positive end expiratory pressure to patient specific respiratory mechanics—seems to be more important than the specific titration method. This has important implications for implementation: individualised positive end expiratory pressure can be adopted using driving pressure guided, compliance guided, or imaging guided approaches on the basis of available expertise and equipment. By contrast, fixed positive end expiratory pressure strategies yield inconsistent results, as shown by the PROVHILO trial, in which fixed high positive end expiratory pressure (12 cm H₂O) failed to reduce PPCs.⁵

However, positive end expiratory pressure strategies—particularly at higher pressure levels—are associated with increased hypotension and vasopressor use,^{5 51} which should be weighed against pulmonary benefits, particularly in haemodynamically unstable patients. Additionally, individualised positive end expiratory pressure titration requires additional time, expertise in interpreting respiratory mechanics, and appropriate monitoring (real time respiratory mechanics display). It should be applied as part of a comprehensive lung protective strategy in selected patients at high risk. For patients at lower risk or in resource limited settings, composite lung protective strategies provide a pragmatic alternative.

Physiotherapy

Early mobilisation represents a key component of enhanced recovery after surgery protocols. It was the only subtype among physiotherapy interventions that achieved moderate certainty evidence for PPC prevention and should be encouraged as part of routine perioperative care. However, optimal implementation in terms of frequency, duration, and intensity remains unclear and warrants further research.

Breathing exercises aim to prevent PPCs through maintaining alveolar recruitment, enhancing respiratory muscle strength, and promoting secretion clearance. Few previous meta-analyses have specifically evaluated breathing exercises in abdominal surgery; the most recent systematic review pooled only two trials and reported protective effects consistent with our expanded 16 trial analysis.²⁹¹ Evidence certainty remains low primarily owing to inherent challenges in masking behavioural interventions and variability in exercise modalities, implementation timing, and supervision approaches. Notably, a large randomised controlled trial by Boden and colleagues showed that a 30 minute preoperative physiotherapy session halved PPC incidence,¹⁵ with the key mechanism being patient empowerment to self-initiate breathing exercises immediately on awakening—substantially earlier than conventional ward based physiotherapy. This suggests that timing of initiation may be critical. Our ongoing multicentre trial (ChiCTR2100047325) is designed to evaluate whether directly supervised breathing exercises after extubation can enhance outcomes.

Analgesia

Analgesia shows benefit for PPC prevention, with the epidural analgesia subtype showing consistent protective effects. This finding is consistent with previous expert opinion that epidural analgesia improves pulmonary outcomes.¹¹ Physiologically, effective pain control facilitates deep breathing, early mobilisation, and effective coughing. These findings support guideline recommendations for optimised perioperative analgesia, including thoracic epidural analgesia for open abdominal surgery.²⁹² The associated sympathetic blockade related hypotension, however, necessitates careful haemodynamic management.

Nutrition

Nutritional optimisation is integral to enhanced recovery after surgery protocols,^{293 294} although previous systematic reviews showed conflicting evidence on PPC prevention.^{295 296 297} Our analysis showed moderate certainty evidence of benefit at the intervention type level, although only supplemental perioperative nutrition support showed benefit among subtypes. This likely reflects inadequate sample sizes limiting statistical power for individual subtypes, substantial protocol heterogeneity (formulation, timing, delivery routes), and, critically, lack of nutritional risk stratification. Theoretically, nutrition interventions may prevent PPCs through maintaining immune competence to resist respiratory tract infections, preserving respiratory muscle strength, and reducing systemic inflammation that can exacerbate pulmonary injury. Future trials should examine specific nutrition strategies in populations stratified by baseline nutritional status and surgical complexity.

Challenging current practice: no evidence of benefit (moderate certainty)

Previous meta-analyses across mixed surgical populations suggested that goal directed haemodynamic therapy reduces PPCs, in contrast to our null finding.^{21 298} The OPTIMISE II trial showed that cardiac output guided therapy failed to reduce infections but increased acute cardiac events,²² and individualised or intensive blood pressure management similarly showed no PPC benefit.^{230 231} Recent international consensus now recommends against use of routine goal directed haemodynamic therapy in major elective abdominal surgery, corroborating our finding.²⁹⁹ Similarly, although excessive fluid administration may impair pulmonary function through interstitial oedema, the RELIEF trial found that restrictive strategies increased acute kidney injury without reducing pneumonia.²⁶³ These null findings likely reflect distinct pathophysiology: PPCs after abdominal surgery arise predominantly from mechanical respiratory impairment (diaphragmatic dysfunction, pain related splinting, and atelectasis) rather than from tissue hypoperfusion or fluid overload, which haemodynamic and fluid optimisation strategies target.

Bi-level positive airway pressure provides dual level pressure support to recruit atelectatic alveoli and maintain functional residual capacity, theoretically benefiting obese patients. However, our analysis failed to show a benefit in this population. This contrasts with a recent meta-analysis reporting that bi-level positive airway pressure reduced atelectasis,³⁰⁰ although the inclusion of observational studies may have introduced bias and overestimated treatment effects. Given the multifactorial pathogenesis of PPCs, bi-level positive airway pressure—primarily targeting ventilatory mechanics without mitigating underlying inflammatory and clearance impairments—may provide insufficient protection in obese patients with heightened susceptibility to atelectasis and hypoxaemia.

Drain omission: uncertain safety signal

Contemporary enhanced recovery after surgery protocols often recommend against routine prophylactic drainage to reduce infection risk and facilitate early mobilisation.^{20 301 302 303} However, our analysis observed a signal suggesting that complete drain omission might be associated with increased PPC risk, a finding that persisted in sensitivity analysis. The mechanism may involve retained intra-abdominal fluid collections facilitating bacterial translocation and systemic inflammatory response or delayed recognition of complications, including anastomotic leaks or bleeding that could progress to respiratory compromise.^{304 305} However, this observation requires cautious interpretation. The small sample size increases vulnerability to type I error, and the low certainty of evidence precludes definitive conclusions. Additionally, the pooled analysis encompassed heterogeneous procedures—hepatic resection, pancreaticoduodenectomy, and ovarian cytoreductive surgery—each with distinct baseline PPC risks and drainage indications, limiting the generalisability of these pooled estimates to specific surgical contexts.

This finding should be considered hypothesis generating rather than practice changing. We do not recommend altering evidence based enhanced recovery after surgery drainage protocols. Future large, well designed, procedure specific trials are needed for confirmation.

Care bundle for PPC prevention

Care bundles, collections of evidence based practices implemented together, have shown effectiveness in preventing catheter related complications,³⁰⁶ surgical site infections,³⁰⁷ and ventilator associated pneumonia.³⁰⁸ The underlying principle is that coordinated delivery of multiple complementary interventions achieves greater risk reduction than implementation of individual practices in isolation, reflecting synergistic effects when evidence based care is systematically organised.

Application of care bundles to prevent PPCs has received limited attention. A Delphi consensus process proposed bundle components, including preoperative exercise training, intraoperative lung protective ventilation, and postoperative respiratory interventions, although acknowledging variability in the supporting evidence quality of individual elements.¹⁴ Quality improvement initiatives such as the US “I COUGH” and the UK “ERAS^+” showed reductions in PPCs,^{294 309} but both used before-after study designs without randomisation, potentially introducing selection bias and temporal confounding. These methodological limitations highlight the need for systematic evaluation of the certainty of evidence underlying bundle components.

Our systematic evaluation using GRADE and trial sequential analysis reveals a striking gap between evidence quality and clinical practice. Bundle development should prioritise interventions according to evidence hierarchy rather than clinical tradition. An evidence stratified approach could anchor bundles around high certainty interventions, incorporate moderate certainty beneficial interventions as secondary components, and remove interventions lacking demonstrated benefit to avoid diluting effectiveness and imposing unnecessary implementation burden.

An evidence stratified bundle design would allow dynamic updates as new evidence accumulates, ensuring that recommendations reflect the current best evidence rather than historical practice. Whether this evidence informed approach improves outcomes compared with existing pragmatic bundles, and whether targeting fewer, higher certainty interventions enhances the fidelity of implementation while maintaining clinical benefit, requires evaluation in appropriately powered trials.

Implications for future research

Our review establishes a clear evidence hierarchy, but future research should focus on further refining this evidence through well designed platform trials, and clinicians should use these findings to inform—rather than replace—clinical judgement when adopting evidence informed perioperative strategies. Priority randomised controlled trials should target interventions that showed treatment signals in our meta-analyses but for which evidence remains insufficient to establish definitive effectiveness. Importantly, further effectiveness trials of goal directed haemodynamic therapy, targeted blood pressure management, restrictive fluid strategies, and bi-level positive airway pressure for PPC prevention seem to be unwarranted. Resources would be better directed towards de-implementation research exploring barriers to discontinuing these practices lacking evidence of benefit in clinical protocols.

Critical knowledge gaps include optimal bundle composition for different surgical populations, long term pulmonary outcomes beyond 30 days, cost effectiveness, and the need for standardised PPC definitions. The StEP-COMPAC consensus offers more precise diagnostic criteria based on international standards, focuses on pathophysiologically related complications, and incorporates severity grading.³¹⁰

Strengths and limitations of study

This systematic review builds on a landmark analysis by Odor and colleagues (2020), which examined mixed surgical cohorts but identified no interventions supported by high certainty evidence.²¹ By focusing exclusively on abdominal surgery, our review achieved greater clinical homogeneity, enabling identification of high certainty benefit for selected interventions.

Additional methodological strengths include the largest evidence base to date (255 trials) and a comprehensive, up-to-date search strategy. We applied a granular framework encompassing 39 intervention subtypes, allowing detailed analysis of strategies targeting distinct perioperative mechanisms.

However, several limitations should be acknowledged. Firstly, we did not search the Cochrane CENTRAL database owing to resource constraints. Nevertheless, CENTRAL’s substantial overlap with MEDLINE, Embase, and Web of Science, combined with more than 22 000 records screened, limit the likelihood of missing key trials. Secondly, although small trials (≤100 participants) were included, we applied safeguards—including GRADE downgrading for imprecision, trial sequential analysis to assess cumulative evidence strength, and sensitivity analyses excluding studies at high risk of bias—to minimise the risk of overestimation.

Heterogeneity across interventions, control practices, and PPC definitions posed additional challenges. Despite granular classification into 39 subtypes, within category heterogeneity persists—for example, individualised positive end expiratory pressure trials used different titration methods, and breathing exercise protocols varied in modality and timing. Additional subgroup analyses by these characteristics were not feasible owing to insufficient trial numbers. Accordingly, pooled estimates should be interpreted as average effects across diverse protocols within each category rather than definitive estimates for any specific protocol. Although we prioritised EPCO aligned endpoints and used conservative outcome rules to prevent double counting, residual variability likely influenced effect estimates. Furthermore, atelectasis as an endpoint has inherent limitations, as mild cases are common postoperatively and may be self-limiting. To mitigate this, we excluded studies assessing only early postoperative atelectasis without clinical significance. Additionally, generalisability is limited by frequent exclusion of patients at high risk (for example, those with severe comorbidities) and sparse reporting of equity related variables such as ethnicity or socioeconomic status. Safety outcomes, cost effectiveness, acceptability to patients, and feasibility of implementation were beyond this review’s scope and warrant further evaluation.

Lastly, we did not do network meta-analysis, as our aim was to assess interventions versus standard care rather than to rank alternatives. Many interventions target distinct perioperative phases and are complementary, making indirect comparisons less clinically meaningful.

Conclusion

This systematic review of interventions for preventing PPCs in abdominal surgery established high certainty evidence for low FiO₂. Moderate certainty evidence also supported lung protective ventilation, physiotherapy, analgesic techniques, and nutrition. Despite its widespread clinical use, goal directed haemodynamic therapy showed no evidence of benefit for PPC prevention, and nor did targeted blood pressure management, restrictive fluid therapy, or bi-level positive airway pressure. These findings do not replace clinical judgment but offer a structured evidence base to support evidence stratified perioperative care bundles that prioritise proven interventions and reconsider practices lacking evidence of benefit. Translation into routine practice will require pragmatic evaluation in real world settings, including large scale platform trials focused on identifying barriers and facilitators to implementation.

What is already known on this topic

• Postoperative pulmonary complications (PPCs) occur frequently after abdominal surgery and are associated with significant increases in morbidity, mortality, and healthcare costs

• Many non-drug interventions are used to prevent PPCs, but their effectiveness is uncertain

• A previous comprehensive systematic review of non-cardiac surgery found no interventions supported by high certainty evidence

What this study adds

• This synthesis of 255 trials identified low FiO₂ as the only intervention with high certainty evidence for PPC prevention in abdominal surgery

• Moderate certainty evidence showed benefits for lung protective ventilation, physiotherapy, analgesic techniques, and nutrition interventions

• Goal directed haemodynamic therapy, targeted blood pressure management, restrictive fluid therapy, and postoperative bi-level positive airway pressure showed no evidence of benefit

Web extra.

Extra material supplied by authors

Web appendix: Supplementary materials

Ethics statements

Ethical approval

Not required.

Data availability statement

All data are freely available in the supplementary file. The data extraction form template and R code for meta-analysis are available from the corresponding author on reasonable request.