Systematic Review of Comparative Studies Evaluating Hyperbaric Oxygen Therapy on Skin Flap and Graft Surgical Indications

Abstract

Background:

This systematic review aims to determine whether pre- and postoperative hyperbaric oxygen therapy (HBOT) incurs any benefits or harms in soft tissue injuries (STIs) considered for skin flap or graft surgery.

Methods:

We included adequately powered clinical trials that enrolled patients with STIs serious enough for skin flaps or grafts; evaluated HBOT (≥1.5 ATA) versus control, placebo, or sham procedures; and analyzed survival/healing and complication rates. MEDLINE, Google Scholar, and Embase were searched for relevant literature through June 30, 2024. The Cochrane risk of bias (RoB 2) and the ROBINS-I tools assessed RoB. Meta-analysis and random effects models analyzed randomized controlled trial (RCTs) evaluating flap/graft survival rates within 4 weeks of HBOT (2-2.5 ATA). The GRADE approach determined the evidence recommendations.

Results:

Among 743 records screened, 45 were assessed; 25 reports from 24 studies (13 RCTs; 11 non-RCTs; 2246 patients) were included. RCT RoB ranged from low RoB (1 trial) to high (2 trials). Nonrandomized trials had moderate RoB (n = 5, 45%), whereas 6 had serious RoB. Four RCTs showed a large, horizontal overall effect size (log odds ratio, 1.045; standard error, 0.3104; z, 3.3.67; P < 0.001; 95% confidence interval, 0.44–1.65). Evidence levels were very low to moderate, with 11 strong and 13 conditional recommendations, including a strong recommendation to use HBOT to heal flaps/grafts in STI/trauma wounds.

Conclusions:

A strong recommendation for pre- and postoperative HBOT is warranted to mitigate the need for flap/graft surgery and heal flaps and grafts, given the potentially life- and/or limb-threatening harms that could otherwise occur.

Takeaways

Question: Does hyperbaric oxygen therapy (HBOT) incur any benefits or harms when used either preoperatively or postoperatively for severe soft tissue injuries (STIs) that are traditional candidates for skin flap or graft surgery?

Findings: The GRADE approach assessed 26 outcomes/subpopulations in 24 comparative studies; evidence levels ranged from very low to moderate. We strongly recommend HBOT to heal flap/grafts in STIs/trauma wounds because of the potential limb- and life-threatening harms that could occur if HBOT is not provided.

Meaning: HBOT incurs a beneficial effect in saving lives and limbs in patients with STIs serious enough to consider skin flaps/grafts.

INTRODUCTION

Severe soft tissue injuries (STIs), including severe trauma wounds such as open fractures, crush injuries, and combined musculoskeletal-vascular injuries, as well as advanced chronic ulcers requiring surgical management, are serious life- and limb-threatening (LLT) injuries. Coinfection and poor vascular supply result in amputation rates as high as 60% and mortality rates as high as 48%.^1–6

Early reconstruction of severe STIs with various types of tissue transfers or flaps and skin grafts has been the standard, reducing amputation rates to 5%–7% at experienced surgical centers.^6–14 However, the use of various tissue flaps and skin grafts to cover severe STIs is still associated with significant morbidity,⁶ and postoperative infection rates in severe open fractures have remained relatively unchanged (25% versus 28%) after 20 years.¹²

Hyperbaric oxygen therapy (HBOT) has been used in jeopardized flaps and grafts since the 1960s to increase tissue oxygen levels, reduce ischemia-reperfusion injury, prevent and/or manage infection, and facilitate wound healing.^15–19 During HBOT, the patient is placed in a monoplace or multiplace hyperbaric chamber, usually for 90–120 minutes, and is exposed to an increased atmospheric pressure while breathing 100% oxygen.¹⁵ The Undersea and Hyperbaric Medical Society (UHMS) recognizes 15 HBOT indications, 7 of which are related to STIs, including arterial insufficiency for problem wounds, compromised flaps and grafts, acute traumatic ischemia, delayed soft tissue radiation injuries, necrotizing soft tissue infections, refractory osteomyelitis, and thermal burns.¹⁵

Previous systematic reviews have analyzed the comparative effectiveness of HBOT to standard surgical/postoperative care in complex acute surgical/trauma wounds, including flaps and grafts.^16–19 Although these systematic reviews suggest that HBOT facilitates graft take, wound healing, and limb salvage, the level of evidence is low to moderate, and high-quality randomized controlled trials (RCTs) with a low risk of bias (RoB) are lacking. Meta-analysis of the effect of HBOT on severe STIs is difficult, given the heterogeneity of trial design and the diverse underlying causal factors in severe injuries.^18,19 A 2013 Cochrane review found that only 2 small RCTs, both with a high RoB, demonstrated the beneficial effect of HBOT on skin grafting.¹⁹

HBOT has traditionally been applied postoperatively in cases of jeopardized flaps and grafts, an indication that most payers in the US cover.^6,20,21 As a result of data suggesting that HBOT can mitigate ischemia-reperfusion injury,¹⁵ recent clinical studies have evaluated the preoperative use of HBOT.^6,20,21 Given that the evidence from previous systematic reviews is more than 10 years old and does not consider preoperative HBOT,^16–19 the objective of this current systematic review of comparative studies is to determine whether HBOT incurs any benefits or harms when used either preoperatively or postoperatively on severe STIs that are traditional candidates for skin flap or graft surgery.

MATERIALS AND METHODS

Study Design and Selection Process

The study protocol details were registered on June 29, 2024, on PROSPERO (no. CRD42024559661). The review began on July 1, 2024. The original focus was to assess the effect of HBOT on survival rates of skin grafts (defined as a piece of healthy skin that is removed from a patient’s donor site and transferred without its blood supply to a recipient site on the same patient) and skin flaps (defined as a section of skin, including the underlying fat, which, with its own blood supply, is transferred to another site). Graft/flap failure was defined as the death of the mobilized tissue or failure of the tissue to be incorporated into the recipient bed. The protocol was amended on August 5, 2024, to expand the research question to include preoperative HBOT to heal severe STIs (defined as injuries to the muscles, ligaments, and tendons) that would otherwise be candidates for skin flaps/grafts, to mitigate the need for skin flaps/grafts, and to reduce potential postoperative complications, as well as the postoperative use of HBOT to heal flaps/grafts.^22,23 The review was completed on February 21, 2025, and adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 Statement and Checklist. All study data are provided in this article.

We included studies that enrolled patients who had STIs serious enough to be candidates for skin flaps/grafts and who received HBOT (defined as ≥1.5 ATA administered in a hyperbaric chamber).¹⁵ All comparative studies (retrospective/prospective, RCTs, and nonrandomized) were considered, provided that the study population was human and that the setting was clinical. Eligible comparator interventions included controls (standard of care only) or placebo procedures, for which patients were placed in a hyperbaric chamber, but they did not receive a therapeutic pressure of at least 1.5 ATA. Alternative sham procedures included administering an initial but momentary pressure of at least 1.5 ATA, which was quickly decreased to below the therapeutic range to provide the sensation of receiving HBOT. Studies that compared HBOT with a different advanced therapy were excluded. Studies had to be powered to detect statistical significance in outcomes or have a sufficient effect size when power calculations were not reported. Outcomes data had to be clear, complete (with n values reported), and correspond to the assessments used to determine flap/graft survival and wound healing rates, reduced complication rates, and/or reduced need for additional/major surgery.

MEDLINE, Google Scholar, and Embase were searched for literature through June 30, 2024. (See appendix, Supplemental Digital Content 1, which displays the detailed literature search strategy, https://links.lww.com/PRSGO/E865.)

Full-text articles of potentially eligible studies and those that did not have enough information in their titles/abstracts to determine their eligibility were retrieved. Two authors screened and selected potentially eligible articles. For any doubts about inclusion, all 4 authors determined eligibility. One author extracted the following data from all studies and compiled them into an Excel spreadsheet: study year, country, trial design, indications, group treatment allocations, treatment duration, follow-up duration, relevant primary/secondary endpoint(s), number of patients, patient age, comorbidities, attrition rate, primary/secondary outcomes, and statistical analysis.

Risk of Bias

Two authors assessed the RoB of RCTs using the Cochrane RoB tool for RCTs (RoB 2).²⁴ They assessed RoB of nonrandomized studies using the RoB in Nonrandomized Studies of Interventions (ROBINS-I) tool.²⁵ If there was no consensus, then the other authors were consulted to make the final assessment.

Meta-analysis

Meta-analysis was performed on RCTs comparing flap survival rates of any kind of flaps that were treated with HBOT (at 2–2.5 ATA) or a control for up to 4 weeks. Meta-analysis was conducted using SPSS (IBM, Armonk, NY) with survival rates as the binary outcome. Random-effects models were used to analyze statistical heterogeneity, with inverse-variance weights applied to both within- and between-study variance. The log odds ratio (OR) measured the effect size. Residual maximum likelihood was the estimation method; there was no standard error (SE) adjustment for the general meta-analysis. Trim-and-fill analysis was performed with a linear estimator imputing from the left side. For model iteration, truncated Knapp–Hartung adjustment was performed. A forest plot, funnel plot, and L’Abbe plot graphically depicted the results of the meta-analysis.

Certainty of Outcomes

One author assessed the certainty of outcomes, based on previously described methodology.²⁶ A binary score (yes/no) was applied if blinding occurred; detailed allocation concealment information was reported; the trial was powered at a minimum of 80%; an intention-to-treat analysis was conducted on all randomized subjects; and the attrition rate group difference was at least 20%. A second author verified the assessment; if there were any disagreements, the other authors were consulted to reach a consensus. If each study characteristic was scored with a “yes,” then there was high certainty of outcomes. If at least 50% of the characteristics were scored “yes,” then there was moderate certainty. If fewer than 50% were “yes,” then there was low certainty.²⁶

GRADE Assessment

Two authors used the GRADE (Grading of Recommendations, Assessment, Development, and Evaluations) approach to consider trial data and RoB analysis to assess the quality of each reviewed outcome from high (ie, there is confidence in the effect estimate and further research is unlikely to affect this confidence) to very low.^27,28 Evidence quality was upgraded if there was a large magnitude of effect, dose response, or plausible confounding that reduced the effect or suggested a false effect. Both authors had to agree to downgrade the evidence based on less than 300 events, less than 400 participants, an underpowered design, considerable RoB or publication bias, or if the 95% confidence intervals (CIs, or alternative estimate of precision) implied little or no effect and appreciable benefit/harm.

An overall evidence recommendation (strong versus weak) was determined. A weak recommendation indicated that engaging in a shared patient-physician decision-making process is essential. A strong recommendation suggested that it is not usually necessary to present an alternative treatment option. When low- or very low-quality evidence was determined, a strong recommendation could still be given when the evidence still implied there was benefit in an LLT situation.^27,28

RESULTS

The literature search resulted in 743 records screened for potential inclusion. The authors sought retrieval of 46 studies, 45 of which were retrievable and assessed for eligibility. Twenty-five reports from 24 studies were included, including 14 publications for 13 RCTs and 11 publications for 11 non-RCTs.^{1,6,21,29–50} Figure 1 depicts the PRISMA flow diagram. (See appendix, Supplemental Digital Content 2, which displays the list of publications screened but ultimately excluded, with reasons for exclusion [n = 20], https://links.lww.com/PRSGO/E866.) Among the studies included, 20 countries were represented and 2246 patients were evaluated, including 1038 patients who received HBOT. Wound types were very heterogeneous, but the majority of studies evaluated nonsurgical, traumatic wounds (13 of 24, 54%). Most studies (18 of 24, 75%) and all RCTs had statistically significantly different healing rates in favor of HBOT (P < 0.05). Table 1 provides an overall summary of the studies included.

Fig. 1.

PRISMA flow diagram. SSTI, severe soft tissue injury.

Table 1.

Summary of Publications Included in This Review

	RCTs (n = 13), n (%)	Nonrandomized (n = 11), n (%)	Total (n = 24), n (%)
Total no. countries	13	10	20
Total no. patients	1018	1228	2246
Total no. HBOT patients	497	541	1038
No. blinded (not open) studies	3 (23)	0 (0)	2 (8)
No. double-blinded studies	1 (8)	0 (0)	1 (4)
No. retrospective studies	0 (0)	6 (55)	6 (25)
No. analyzing trauma/fractures wounds	9 (69)	4 (36)	13 (54)
No. analyzing crush injuries	3 (23)	2 (18)	5 (21)
No. analyzing surgically managed DFUs	1 (8)	1 (9)	2 (8)
No. analyzing pressure injuries	1 (8)	0 (0)	1 (4)
No. analyzing burns	1 (8)	1 (9)	2 (8)
No. analyzing radiation injuries	0 (0)	1 (9)	1 (4)
No. analyzing surgical wounds/excisions	1 (8)	5 (45)	6 (25)
No. analyzing hypospadias	0 (0)	4 (36)	4 (17)
No. with statistically significantly different (P < 0.05) healing rates in favor of HBOT	13 (100)	5 (45)	18 (75)

Supplemental Digital Content 3 summarizes RCT study design characteristics. (See appendix, Supplemental Digital Content 3, which displays the study design characteristics of RCTs [n = 13], https://links.lww.com/PRSGO/E867.) Table 2 summarizes the patient characteristics. Supplemental Digital Content 4 lists the RCT analyses. (See appendix, Supplemental Digital Content 4, which displays the summary of analysis and results of RCTs [n = 13], https://links.lww.com/PRSGO/E868.) Supplemental Digital Content 5 summarizes study design characteristics of the 11 nonrandomized trials; Table 3 lists their patient characteristics. (See appendix, Supplemental Digital Content 5, which displays the study design characteristics of nonrandomized, comparative trials [n = 11], https://links.lww.com/PRSGO/E869.) Supplemental Digital Content 6 summarizes their outcomes. (See appendix, Supplemental Digital Content 6, which displays the summary of analysis and results of nonrandomized, comparative trials [n = 11], https://links.lww.com/PRSGO/E870.)

Table 2.

Patient Characteristics of RCTs (n = 13)

Authors	No. Patients in HBOT (Group 1) Versus Comparator Group(s) (Groups 2 and 3)	Mean Age (SD), y, in HBOT (Group 1) Versus Comparator Group(s) (Groups 2 and 3)	Comorbidities, n (%), in HBOT (Group 1) Versus Comparator Group(s) (Groups 2 and 3)	Attrition Rate, %, in HBOT (Group 1) Versus Comparator Group(s) (Groups 2 and 3)
Bouachour et al29	18 versus 18	45.8 (16.1) versus 51.5 (20.9)	• Diabetes: 2 (11) versus 4 (22) • Prior neurological deficit: 1 (6) versus 1 (6)	0 versus 0
Chen et al30	40 versus 40	30.8 (8.6) versus 28.6 (6.5)	(–)	0 versus 0
Chen et al31	68 versus 52	(–)	(–)	(–)
Cheng et al32	32 versus 32	38.2 (16.3) versus 37.1 (17.2)	(–)*	0 versus 0
Duzgun et al33	50 versus 50	58.1 (11) versus 63.3 (9.2)	• Hypertension: 32 (64) versus 28 (56) • BMI >30 kg/m2: 40 (80) versus >30 kg/m2: 23 (46) • Smoker: 36 (72)† versus 20 (40) • LDL ≥160 mg/dL, TG ≥180 mg/dL, cholesterol ≥200 mg/dL: 31 (62) versus 27 (54) • Insulin dependent: 41 (82) versus 45 (90	0 versus 0
Huang et al34	40 in each of 3 groups	32 in group 1 versus 32 in group 2 versus 30 in group 3	(–)	0 in each of 3 groups
Jiang and Wei35	32 versus 32	31 (8) versus 29 (7)	(–)	0 versus 0
Liu and Yang36	30 versus 30	36 versus 35	• Diabetes: 3/30 (10) versus 0/30 (0) • Hypertension: 2/30 (7) versus 3/30 (10) • Secondary anemia: 2/30 (7) versus 1/30 (3)	0 versus 0
Millar et al6,37	60 versus 60	40 (31.0–55.5)* versus 40 (27.0–53.0)*	• Median (IQR) BMI, kg/m2: 26.5 (5.7) versus 25.2 (5.9) • Smoker: 18 (30) versus 15 (26) • Diabetes: 2 (3) versus 2 (3) • Hypertension: 6 (10) versus 9 (15) • Impaired ambulation: 2 (3) versus 7 (12) • Autoimmune disease: 4 (7) versus 2 (3) • PVD: 1 (2) versus 1 (2)	• 2 wk: 3 versus 2 • 12 mo: 12 versus 10 • 24 mo; 42 versus 48
Perrins38	24 versus 24	(–)	(–)	0 versus 0
Raj et al39	30 versus 30	• 18–40: 5 (17%) versus 18–40: 6 (20%) • 40–60: 20 (67%) versus 40–60: 18 (60%) • >60: 5 (17%) versus >60: 6 (20%)	(–)	0 versus 0
Xu et al40	37 versus 37	29.2 versus 21.5	• Diabetes: 0 versus 0	0 versus 0
Yan et al41	36 versus 36	25.3 versus 24.4	(–)‡	0 versus 0

^*Although comorbidity data were not reported, the authors noted that there were statistically significant differences in sex, age, disease type, and condition between the study groups (P < 0.05).

^†The HBOT group had more severe cases, with significantly more patients being male, obese, and having a history of tobacco use (P < 0.05).

^‡Although authors did not report comorbidity data and reported very limited data on patients’ characteristics, they noted that the clinical data of the patients’ conditions were comparable between groups (P > 0.05).

BMI, body mass index; IQR, interquartile range; LDL, low-density lipoprotein; PVD, peripheral vascular disease; TG, triglyceride; (–), missing data.

Table 3.

Patient Characteristics of Nonrandomized, Comparative Trials (n = 11)

Authors	No. Patients in HBOT Versus Comparator Group	Mean Age (SD), y, in HBOT Versus Comparator Group	Comorbidities: n (%) in HBOT Versus Comparator Group	Attrition Rate, %, in HBOT Versus Comparator Group
Bush and Snodgrass21	32 versus 25	12 versus 11	Postpuberty: 10 (31%) versus 7 (28%) Mean no. prior failed repairs: 3.5 versus 3.5 (P = 0.001)	0 versus 0
Chang et al42	49 versus 33	(–)	(–)	0 versus 12
Chiang et al43	38 versus 15	22 (1) versus 23.4 (1)	None. All patients were otherwise healthy	0 versus 0
Lee et al44	21 versus 34	44.8 (17) versus 42.6 (14)	Hypertension: 5 (24%) versus 3 (9%) Diabetes: 0 versus 1 (3%) Current smoker: 7 (33%) versus 9 (26%)	0 versus 0
Lei et al45	36 versus 21	1–4 y versus 1–4 y	(–)	0 versus 0
Lopes et al46	7 (6 had flaps) versus 9 (7 had flaps)	70 (42–85) versus NR, but reported to be similar to the HBOT group	Diabetes: 4/7 (57%) versus similar to the HBOT group Overweight: 4/7 (57%) versus similar to the HBOT group Hypertension: 1/7 (14%) versus similar to the HBOT group	0 versus 0
Marra et al47	11 versus 10	38.4 (11.3) versus 31 (5.9)	(–)	0 versus 0
Meier et al48	15 in each of 3 groups	Median (IQR): 55 (47–61) in group 1; 54 (47–61) in group 2; 58 (50–64) in group 3	Median (IQR) BMI, kg/m2: 26 (24–28) versus 27 (26–30) in group 2; 26 (24–29) in group 3 Never smoked: 7 (47%) versus 7 (47%) in group 2; 10 (67%) in group 3 Median (IQR) no. risk factors*: 1(0–2) versus 1 (1–2) in group 2; 2 (1–2) in group 3	0 in all groups
Neheman et al49	99 versus 289	2.7 (1.4) versus 4 (2.2)	Mean (SD) weight, kg: 14.3 (2.3) versus 3.2 (13.7) Mean (SD) no. prior operations/patient: 4.2 (2.5) versus 3.1 (2.7), P = 0.38 Mean (SD) no. complications/patient: 5.3 (2.1) versus 1.7 (0.9), P < 0.0001	36 versus 0
Roje et al50	16 versus 13	Median 29 versus (–)	(–)	0 versus 0
Yamada et al1	49 versus 33	54 (20) versus 55 (29)	Smoker: 6 (38%) versus 1 (8%) Hypertension: 6 (38%) versus 4 (31%) Hyperlipidemia: 2 (12%) versus 1 (8%) CAD: 1 (6%) versus 0 Arrhythmia: 1 (6%) versus 0 Cerebrovascular disease: 0 versus 1 (8%)	0 versus 0

^*Diabetes, deep vein thrombosis, BMI ≥25, antiplatelet therapy, thrombosis, hypertension, anticoagulants, pulmonary embolism, and smoking history.

BMI, body mass index; CAD, coronary artery disease; NR, not reported; (–), missing data.

Table 4 depicts the RoB 2 assessment of each RCT, whereas Table 5 provides an overall summary of each RoB 2 domain and overall RoB. One RCT (8%) had an overall low RoB; it was the only trial that had a low RoB for the selection of reported results.²⁹ Most RCTs had some concerns or high risk for the randomization, assignment to intervention, and deviations from the intended intervention domains. Ten (77%) RCTs had some concerns overall; 2 (15%) had a high overall RoB.

Table 4.

RoB 2 Assessment of RCTs (n = 13)

Table 5.

Summary of Number (%) of RCTs (n = 13) at Each Risk Level for Each RoB 2 Domain and Overall RoB

Trial Risk Level	Overall RoB, n (%)	Randomization, n (%)	Assignment to Intervention, n (%)	Deviations From the Intended Interventions, n (%)	Missing Outcome Data, n (%)	Outcome Measurements, n (%)	Selection of Reported Results, n (%)
Low	1 (8)	2 (15)	3 (23)	5 (38)	13 (100)	12 (92)	1 (8)
Some concerns	10 (77)	10 (77)	10 (77)	7 (54)	0 (0)	0 (0)	11 (84)
High	2 (15)	1 (8)	0 (0)	1 (8)	0 (0)	1 (8)	1 (8)

Table 6 summarizes the certainty of RCT outcomes. Trials had predominantly low certainty of outcomes (n = 10, 77%). Three trials (23%) had moderate certainty. Only 1 trial (8%) blinded patients and study assessors. Two trials (15%) had allocation concealment. Almost all trials (n = 12, 92%) used an intention-to-treat analysis for primary endpoints. Only 1 trial reported a power calculation of at least 80%. All 13 trials had a group difference in attrition rate of less than 20%.

Table 6.

Certainty of Outcomes Assessment of RCTs (n = 13)

Authors, Year	Patients Blinded	Study Assessors Blinded	Allocation Concealment	ITT Analysis for Primary Endpoints	Power Calculation ≥80%	Attrition Rate <20% Group Difference	Overall Certainty of Outcomes
Bouachour et al, 199629	Y	Y	N	Y	N	Y	Moderate
Chen et al, 200230	N	N	N	Y	N	Y	Low
Chen et al, 201131	N	N	N	Y	N	Y	Low
Cheng et al, 200832	N	N	Y	Y	N	Y	Low
Duzgun et al, 200833	N	N	N	Y	N	Y	Low
Huang et al, 200534	N	N	N	Y	N	Y	Low
Jiang and Wei, 200435	N	N	N	Y	N	Y	Low
Liu and Yang, 200636	N	N	N	Y	N	Y	Low
Millar et al, 202237	N	Y	Y	N	Y*	Y	Moderate
Perrins, 196738	N	Y	N	Y	N	Y	Moderate
Raj et al, 202339	N	N	N	Y	N	Y	Low
Xu et al, 200040	N	N	N	Y	N	Y	Low
Yan et al, 200941	N	N	N	Y	N	Y	Low

^*Millar et al³⁷ was ultimately underpowered.

ITT, intention-to-treat.

For nonrandomized studies, Table 7 summarizes their individual RoB; Table 8 summarizes their overall collective risk. Nonrandomized trials had moderate (n = 5, 45%) or serious (n = 6, 55%) RoB. All 6 trials with serious RoB had serious confounding bias (Table 8). Trials with moderate RoB were limited predominantly by their biased outcome measurements (n = 10, 91%).

Table 7.

ROBINS-I Assessment of Bias of Nonrandomized, Comparative Studies (n = 11)

Table 8.

Summary of Number (%) of Nonrandomized Comparative Studies (n = 11) With Each Risk Level for Each ROBINS-I Domain and Overall RoB

Trial Risk Level	Overall RoB, n (%)	Confounding Bias, n (%)	Selection Bias, n (%)	Classification of Interventions, n (%)	Deviations From Intended Interventions, n (%)	Missing Data, n (%)	Outcome Measurements, n (%)	Selection of the Reported Results, n (%)
Low	0 (0)	3 (27)	8 (73)	10 (91)	10 (91)	10 (91)	1 (9)	11 (100)
Moderate	5 (45)	2 (18)	2 (18)	1 (9)	1 (9)	0 (0)	10 (91)	0 (0)
Serious	6 (55)	6 (55)	1 (9)	0 (0)	0 (0)	1 (9)	0 (0)	0 (0)
Critical	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)

Four RCTs (31%) were eligible for the overall meta-analysis. There was a large, horizontal overall effect size (log OR = 1.045; SE, 0.3104; z = 3.3.67; P < 0.001; 95% CI = 0.44–1.65). The P value suggests that differences are very unlikely due to chance. There was a fifth study imputed into the trim-and-fill analysis, with a log OR of 0.809 (SE, 0.3474; z = 2.329; P = 0.020; 95% CI = 0.13–1.49). The OR still indicated a relatively large effect size. Figure 2 depicts the forest plot of the random-effects model of RCTs comparing flap survival rates of HBOT to a control through up to 4 weeks of treatment. For heterogeneity analysis, tau² was 0.14, H² was 1.58, and I² was 0.37. Because tau² approaches 0, consistency and homogeneity are indicated. I² represents heterogeneity, which was less than 40%. Overlapping CIs are observed. Figure 3 depicts the funnel plot of RCTs comparing flap survival rates of HBOT to a control during up to 4 weeks of treatment. Figure 4 depicts the L’Abbe plot with random-effects weights of RCTs comparing flap survival rates of HBOT (treatment group) to a control through up to 4 weeks of treatment.

Fig. 2.

Forest plot of a random-effects model of RCTs comparing flap survival rates of HBOT versus control during a treatment period of up to 4 weeks.

Fig. 3.

Funnel plot of RCTs comparing flap survival rates with HBOT versus control during a treatment period of up to 4 weeks.

Fig. 4.

L’Abbe plot with random-effects weights of RCTs comparing flap survival rates with HBOT (treatment group) versus control during a treatment period of up to 4 weeks.

The GRADE approach assessed 4 outcomes and their 26 subpopulations treated with HBOT. Appendix 7 provides the detailed GRADE assessment for each outcome and its subpopulations. (See appendix, Supplemental Digital Content 7, which displays the results of GRADE assessment, https://links.lww.com/PRSGO/E871.) Appendix 8 lists the recommendations for each outcome and subpopulation. (See appendix, Supplemental Digital Content 8, which displays the summary of GRADE recommendations for each outcome and subpopulation, https://links.lww.com/PRSGO/E872.) Evidence levels ranged from very low to moderate. There were 11 strong recommendations for pre- and postoperative HBOT use and 13 conditional recommendations in nonsurgical traumatic wounds, surgical wounds (related to cancer treatment and hypospadias repair), crush injuries, exposed fractures, pressure injuries, burns, and surgically managed Wagner grade 3 or 4 diabetic foot ulcers (DFUs).

We strongly recommend HBOT to reduce postoperative complications in skin flaps and grafts applied to nonsurgical trauma wounds, crush injuries, and exposed fractures, even though the evidence is very low or low, because of the potential LLT harms that could occur if HBOT is not provided.

We conditionally do not recommend HBOT in fingertip crush injuries to heal flaps and grafts because better-designed studies are still needed to be able to make an appropriate and complete assessment. However, HBOT is recommended to improve the healing rates of wounds, grafts, and flaps in all other fingertip STIs, crush injuries, burns, and hypospadias repair, and to reduce the complication rate and need for additional surgery in nonfingertip crush injuries. We strongly do not recommend HBOT to reduce the need for multiple skin grafts in severe burns.

DISCUSSION

Previous systematic reviews have generated limited recommendations for the use of HBOT on skin flaps and grafts, largely due to their research questions focusing on the general topic of acute wounds.^17–19 Consequently, the beneficial effect of HBOT on healing skin flaps and grafts was an outcome but not the focus of the reviews, and only 2 RCTs were revealed to have low-quality evidence directly related to healing rates of skin flaps and grafts.¹⁹ This current systematic review uniquely acknowledges the evolving approaches to soft tissue coverage in surgical and trauma care and considers a more comprehensive, up-to-date, and multifaceted research question that focuses on the use of both preoperative and postoperative HBOT. Outcomes addressed skin flaps and grafts in terms of severity of the STI, including both acute and chronic etiologies that would have historically been candidates for skin flaps and grafts. Importantly, 9 (35%) of the 26 GRADE recommendations determined by this review support the preoperative use of HBOT to help heal skin flaps and grafts in repaired hypospadias, to heal surgically managed (including by flap and graft surgery) Wagner 3 and 4 DFUs, and to mitigate and reduce the need for flaps and grafts in severe exposed fractures and crush injuries (Supplemental Digital Content 7, https://links.lww.com/PRSGO/E871; Supplemental Digital Content 8, https://links.lww.com/PRSGO/E872).

The literature search methodology and inclusion and exclusion criteria eliminated publication bias by searching beyond the obvious terminology strictly related to “flap” and “graft” to focus on the severity of injury. Additionally, foreign language articles were translated, leading to the inclusion of 8 Chinese comparative clinical trials that were previously excluded from Cochrane systematic reviews. More than 1000 patients from 20 geographically dispersed countries received HBOT in our analysis (Tables 2, 6). The majority of the 24 studies evaluated in this review were RCTs (13 of 24), and meta-analysis of 4–5 RCTs demonstrated there was low heterogeneity and a large horizontal effect size with an OR of approximately 8–10 (Figs. 2–4), providing evidence that HBOT improves graft and flap survival. It should be noted, however, that our inclusion criteria meant that a number of other studies could not be included, and consequently, there is a limitation in the meta-analysis interpretation.

Although the conclusions of this systematic review may at first glance be limited by low certainty of RCT outcomes (Table 6), these findings should be understood in the context of designing and reporting RCTs for severe, often LLT wounds. For example, for ethical reasons, these trials limit blinding opportunities.⁵¹ We included trials with a therapeutic HBOT threshold of at least 1.5 ATA, although standard of care is generally at least 2.0 ATA;¹⁵ however, only 1 Chinese trial evaluated used less than 2.0 ATA in pediatric patients only (Supplemental Digital Content 3, https://links.lww.com/PRSGO/E867).⁴¹ There are reporting limitations due to journal word counts and English language barriers. Only 1 RCT reported an adequate power calculation, but that does not imply that other trials did not have a sufficient effect size or statistical significance, which is what the GRADE assessment attempted to capture and interpret in more detail. Despite the low certainty of outcomes of the RCTs, the GRADE assessment revealed up to a moderate level of evidence. Indeed, all RCTs showed statistically significant healing rates favoring HBOT (Supplemental Digital Content 4, https://links.lww.com/PRSGO/E868). This is particularly remarkable given that, when surgery is used for life-threatening, traumatic injuries (the wound types evaluated in the majority of the studies in this review), trial enrollment and implementation are extremely challenging.⁵¹

Given the limitations of designing RCTs evaluating LLT injuries, it is not surprising that none of the outcomes evaluated in this review had high-quality evidence determined by the GRADE assessment. Still, there are 11 strong recommendations for pre- and postoperative HBOT use and 13 conditional recommendations, which altogether strongly support that HBOT is beneficial to improve flap- and graft-related outcomes in severe STIs.

Importantly, there are strong recommendations for the use of preoperative HBOT to promote healing in surgically managed Wagner grade 3 or 4 DFUs, severe crush injuries, and severe exposed fractures, and to reduce the need for major surgical procedures, amputations, and flaps and grafts. Future comparative research should continue to investigate the preoperative use of HBOT in facilitating the healing of severe STIs and reducing the need for skin flaps and grafts compared with postoperative use.

CONCLUSIONS

We strongly recommend using HBOT to (1) heal flaps and grafts in STIs and trauma wounds; (2) reduce postoperative flap and graft complications in nonsurgical wounds, severe crush injuries, surgically managed Wagner grade 3 and 4 DFUs, and exposed fractures; (3) reduce the need for additional or major procedures in severe crush injuries and surgically managed Wagner grade 3 and 4 DFUs; and (4) heal severe crush injuries, severe exposed fractures, and surgically managed Wagner grade 3 and 4 DFUs. We make these recommendations despite the lack of high-quality evidence, given the LLT nature of these conditions. Overall, this systematic review and meta-analysis demonstrates the beneficial effect of HBOT when applied both preoperatively and postoperatively to severe STIs (in particular, nonsurgical trauma wounds), to heal wounds and skin flaps and grafts, and to mitigate the need for skin flaps and grafts in both acute and chronic LLT STIs. We emphasize the beneficial role of HBOT in saving lives and limbs in patients with STIs serious enough to consider skin flaps and grafts.

DISCLOSURES

Marissa J. Carter is a paid consultant to UHMS. Kristen A. Eckert is a paid consultant to Strategic Solutions and UHMS. The other authors have no financial interest to declare in relation to the content of this article. This work was supported by UHMS, which was consulted for medical expertise during this study.