Hyperbaric oxygen therapy Guidance to commissioners

https://www.birmingham.ac.uk/Documents/college-mds/haps/projects/WMCSU/CommissioningPolicies/HBO.pdf

The evidence base

 The primary research studies investigating the efficacy of HBO are remarkable for the consistent poor quality of the published clinical trials as well as the lack of evidence demonstrating significant health benefits. There is a lack of adequate clinical evidence to support the view that HBO therapy is efficacious for any of the indications for which it is being used. The nature of the treatment would suggest that the potential for significant placebo effects is high and because the evidence base consists mostly of uncontrolled case studies, significant overestimation of any observed effect size is highly likely. Published controlled clinical trials are small in number, generally poorly designed or reported, and their results do not justify the recommendations made for the use of HBO, as judged by current standards of robust evidence-based practice. 

In 2012 NHS commissioners find themselves in the unfortunate situation that despite HBO having been used for many years as a therapeutic intervention for a variety of indications, the evidence for its use remains poor and uncertain, and we are no closer to establishing its efficacy, let alone comparative effectiveness or cost-effectiveness. 

Indications for which HBO use was accepted standard practice

Decompression sickness and gas embolism

Spending time at a raised environmental pressure (e.g. during SCUBA diving) results in additional nitrogen

dissolving in the blood and other tissues. If decompression to normal atmospheric pressure occurs too

quickly to allow the dissolved gas to be eliminated by the lungs, gas bubbles may form in the blood or tissues

(this is called decompression sickness or “the bends”), which may cause a variety of symptoms such as skin

itching, joint pain, and fatigue. If these bubbles enter the arterial circulation through the lungs (gas embolism),

they may cause serious neurological injury ranging from visual disturbance to quadriplegia. The term

decompression illness may or may not include gas embolism, which may also be caused by trauma, including

that caused by surgery, mechanical ventilation and haemodialysis.3

Postulated mechanism of action of HBO

Both decompression sickness and gas embolism can be treated by recompression followed by slower

decompression. Hyperoxia additionally increases the diffusion gradient with the embolised gas, enhancing

the movement of the gas into solution.

Treatment methods

Tables have been developed using mathematical models of the ascent process depending on the depth and

duration of descent. However, since individuals can undergo physiological adaptation to work at high

pressure, there is debate about the application of the tables in practice. Recompression with air was standard

practice until 2001, when the Health & Safety Executive approved oxygen recompression (i.e. HBO) as a

result of their work on rates of bubble formation during air and oxygen recompression.1

 The UHMS

recommends treatment with HBO at 2.8 ATA, repeated up to 10 times if symptoms persists in the case of

decompression sickness, or until no further improvement is seen, usually after 5 to 10 treatments (no further

details available).3

Evidence

The highest level of relevant evidence for decompression sickness comprised 15 reports of case series; there

were no controlled studies.1

 A Cochrane review found two RCTs but these did not use appropriate control

groups to assess the efficacy of HBO treatment (one assessed the effect of adding an NSAID to HBO and the

other assessed the effect of helium plus oxygen compared with oxygen alone). The reports on the case

series suggested that many cases were effectively treated by HBO, with 30 to 98% of patients experiencing

full recovery, although the limitations of such studies were recognised (absence of control group, selection

bias, poor reporting).1

Three case series were identified that studied the effect of HBO on gas embolism specifically. These

provided conflicting evidence, but were limited by the relatively small size of the studies, the use of different

treatment schedules and insufficient detail in the reporting. 

Acute CO poisoning

This indication is discussed here because of the perception that HBO treatment is standard care for patients

with acute CO poisoning. However, such treatment is not generally recommended in the UK (see National

guidance below).

Carbon monoxide binds to haemoglobin (forming carboxyhaemoglobin), with an affinity about 200-fold greater

than oxygen, and also increases the affinity of the unoccupied sites for oxygen, thereby reducing the capacity

of the blood to carry oxygen as well as to deliver it.4

 This leads to oxidative stress, cell apoptosis, inflammation and necrosis, neurological and cardiac damage and potentially, death. Neurological and

psychological sequelae may persist or develop days or weeks after acute CO poisoning, in up to 50% of

poisoned patients.5

 These effects are thought to involve mitochondrial dysfunction and lipid peroxidation in

the brain. They may include cognitive, affective and motor disturbances, peripheral neuropathy, hearing loss,

and dementia and psychosis.5

Postulated mechanism of action of HBO

The elimination half-life of carboxyhaemoglobin is 4 to 6 hours in room air, and 80 to 100 minutes with

administration of 100% oxygen at atmospheric pressure (normobaric oxygen, NBO). HBO at 3 ATA

decreases the half-life to about 20 minutes4

 (no primary references provided in article). Furthermore, HBO at

3 ATA increases the oxygen dissolved in the plasma from about 3 ml/L at normal pressure to about 60 ml/L,

which is almost sufficient to supply the requirement of many tissues independent of oxygen bound to

haemoglobin.4

 Whether HBO improves survival or long-term outcomes compared with 100% oxygen at

normal pressure is recognised to be unclear from the evidence (see below).

National and international guidance

In the UK, the National Poisons Information Service (NPIS), on the website Toxbase, stated in 2008 that:

“Treatment with hyperbaric oxygen is not currently recommended, because there is insufficient evidence that

hyperbaric oxygen therapy improves long-term outcomes of people with severe monoxide poisoning,

compared with standard oxygen therapy.”6

NHS Direct, in NHS Choices, states: “There is currently insufficient evidence regarding the long-term

effectiveness of HBO for treating severe cases of CO poisoning. Therefore, standard oxygen therapy [100%

oxygen given through a mask] is usually the recommended treatment option.”

Evidence

A Cochrane review found six RCTs that used clinical outcomes measuring the effect of HBO following acute

CO poisoning compared with NBO.8

 The trials were of varying quality, with a total of 1,361 patients (one of

the trials was available only as an abstract of an interim analysis). Included patients had CO poisoning of

varying severity, and different regimens of HBO and NBO were used. Two trials used “sham” dives for the

control group, exposing patients to NBO in a hyperbaric chamber, and these trials were stated to be doubleblind. Virtually all patients received supplemental oxygen prior to randomisation, but the exact time point after

CO exposure at which patients were treated with HBO or NBO is not clear from the review. Extent of followup was considered to be poor and the risk of bias high, according to the authors of the review.8

 Most of the

RCTs focussed on outcomes determined by non-specific self-reported symptoms. Several of the trials were

terminated prematurely and at least in one, this resulted in lack of power to detect differences between treated

groups.9

 (This trial had two parts: one compared HBO with NBO and the other compared one or two sessions

of HBO added to NBO. At the interim analysis, recovery rates were significantly lower with two sessions of

HBO compared with one session, and the trial was halted. The other part of the trial comparing HBO with

NBO was also halted because no difference was seen between the groups and it was felt to be “futile” to

continue. However, at that point, only 179 patients had been randomised into this part of the trial as opposed

to the 490 that had been calculated would be required to show a significant difference. Therefore, the trial

was significantly underpowered.)

Of the six trials, two found a beneficial effect of HBO for the reduction of neurological sequelae at four to six

weeks after randomisation, while the other four trials found no benefit. A meta-analysis combining results of

all six trials found no evidence of benefit, with an odds ratio for neurological deficits for HBO compared with

NBO of 0.78, 95% CI 0.54 to 1.12. 8

 However, significant methodological and statistical heterogeneity was

found among the trials, which made the result of the meta-analysis less certain. Only one of the two trials

including a “sham” control group found a positive benefit. The authors concluded that additional clinical trials

comparing HBO with NBO are “ethical, warranted and necessary” because of the many limitations of the

currently published trials. 8

A randomised, double-blind, clinical trial is underway to compare the effect of one versus three HBO sessions

on the incidence of cognitive sequelae at six weeks after acute CO poisoning. The study was begun in 2007

and the estimated completion date is May 2013 (ClinicalTrials.gov number NCT00465855). 

Indications for which reasonable evidence indicated no efficacy for HBO

Multiple sclerosis

One systematic review and meta-analysis (an update of an earlier Cochrane review by the same authors)

found 12 RCTs conducted between 1983 and 1987.10 The meta-analysis concluded that there was no

clinically significant benefit from the administration of HBO. Most of the trials tested HBO against placebo, at

pressures between 1.75 and 2.5 ATA daily, in 20 treatments over four weeks.

In March 2003, NICE stated that HBO “should not be used in patients with multiple sclerosis because

research evidence did not show beneficial effects on the course of the condition.” (NICE CG 8, 2003). 

Indications for which the evidence is suggestive and further research is warranted 

Diabetic foot ulcer (Table 3a, p. 16)

Chronic wounds are defined as lesions that take a long time to heal, fail to heal or recur. The most common

chronic wounds in western countries are diabetic foot ulcers. It has been estimated that 15% of patients with

diabetes mellitus suffer foot ulceration (caused by neuropathic and vascular complications), and in these

patients the amputation rate is 15-70 times higher than in the general population.11 Other common chronic

wounds are venous leg ulcers (caused by sustained high venous pressures), arterial leg ulcers (caused by

arterial insufficiency), and pressure ulcers (caused by unrelieved pressure or friction). Treatment of chronic

wounds involves multidisciplinary approaches including treating underlying pathology, systemic treatment

such as nutritional supplements, and local treatment to improve the wound environment, such as dressings

and pressure-relieving mattresses. However, chronic wounds are resistant to treatment and may slowly

progress despite treatment.

In 2011, the Drug and Therapeutics Bulletin (DTB) commented on the incidence of major amputations (at the

ankle or above) related to type 2 diabetes in the UK.12 This incidence rose from 2.0 to 2.7 per 100,000

population between 1996/7 and 2005/6. NICE has advised that in-patients with a diabetic foot problem should

be referred to a multidisciplinary specialist foot team,13 but audit data showed that 20% of hospitals did not

have such a team and only 28% of inpatients with diabetes had their feet checked while in hospital. The DTB

report further cites evidence that showed a two-fold variation between strategic health authorities in England

in major amputation rates in people with type 2 diabetes, and an eight-fold variation between primary care

trusts in major amputation rates related to diabetes.12

Postulated mechanism of action of HBO

HBO therapy is thought to decrease hypoxia of wound tissue, enhancing perfusion, reducing oedema,

promoting collagen production, and angiogenesis.3

 However, the role of oxygen in wound healing is

complex11 and is based largely on in vitro studies. Angiogenesis appears to be stimulated by hypoxia, while

other phases of tissue repair are oxygen-dependent, e.g. fibroblast proliferation and bacterial killing by

macrophages. Therefore, a correct balance between hypoxia in the wound and oxygenation around the

wound appears to be associated with successful healing.11 This makes prediction of the effect of raising

tissue oxygen pressure on wound healing difficult.

Evidence

The primary source evidence, which consists of six published RCTs,14-19 several non-randomised controlled

studies, and a number of case series, has been assessed in various combinations in many systematic

reviews. Subsequent systematic reviews and HTAs have assessed or summarised older reviews, thereby

creating a large body of reports that is based on a small group of primary studies. None of the systematic

reviews analysed all six RCTs. The RCTs are summarised in Table 3a and described below.

The RCTs were published between 1992 and 2010, and varied considerably in quality of design, patient

inclusion criteria, details of treatment regimen, follow-up periods and completeness of reporting. The number

of patients in individual studies ranged from 18 to 100. Included patients had diabetes with foot or lower

extremity ulcers that were described as “chronic” or had failed to heal with standard therapy over specific

periods ranging 4 weeks to 6 months across the studies. One study restricted inclusion to patients with

wounds of Wagner grade 1 to 3, another used Wagner grade 2 to 4.

In all studies, HBO was adjunctive to standard therapy, which was given to all patients. Standard therapy was

described in detail in three studies,14,17,19 as “regular surgical treatment” with a brief description also

mentioning antibiotics in one trial,15 as “conventional medical treatment” with a brief description,18 and as

“daily wound care” with a brief description.16 HBO was compared with either hyperbaric air therapy in a

double-blind design (in two trials14,19) or no adjunctive therapy in an open design (four trials).

HBO pressure varied from 2.2 ATA to 3.0 ATA, the treatment duration of each session from 30 to 90 minutes,

and the number of sessions from 4 sessions over 2 weeks, to 40 sessions over 8 weeks. The main outcomes

measured included:

• reduction in wound size

• number of patients with ulcer healing

• number of major (or proximal) amputations

• number of minor (or distal) amputations 

changes in transcutaneous oxygen tension (TcPO2)

The time points at which these outcomes were measured were not stated in two studies.15,17 One study did

not prespecify a duration of follow-up but reported a mean follow-up period of 92 weeks.16 One study

measured wound size at the end of treatment and after a further four weeks.18 The two double-blind shamcontrolled studies assessed outcomes for up to one year.14,19 (Omission of the time point of assessment

introduces the potential for considerable bias if the treated patients and the control patients have different

follow-up periods.)

Methodological quality of the trials varied, with Jadad scores ranging from 1 to 4, with three trials scoring 2.

(Jadad scores reflect randomisation, blinding and adequate description of trials; 5 is the maximum score.) A

power calculation was reported in one of the trials and intention-to-treat analysis in one other.

Wound size reduction. This was measured in two trials including one of the double-blind sham-controlled

trials.14 In one trial18 (n = 28), the wound size decreased significantly more (by about 20%) during the twoweek treatment period with HBO compared with no treatment, but there was no significant difference at final

follow-up four weeks later. The second trial14 (n = 18) also found a difference in the reduction in wound size at

the end of treatment (compared with sham treatment) but not at final follow-up assessment at 6 months. In

both these trials, wound size decreased more rapidly in the control groups compared with the treated group,

resulting in no difference between the groups at the final assessment.

Proportion of ulcers healed. This outcome was measured in three trials. Significant differences were

reported between the HBO and control groups in all three trials:

5/8 vs. 0/8 at final follow-up at one year (p < 0.05)14

52% vs. 29% of patients after one year (p < 0.05, n = 94)19

66% vs. 0% of patients (p < 0.05, n = 100)16 (in this study no prospectively specified duration of

follow-up was reported)

Major amputation. The incidence of major amputation was reported in five studies.14-17,19 It was measured

at variable times after HBO treatment. In the five individual studies, the numbers of amputations in the HBOand control groups respectively were:

3 vs. 1 at one year (total n in trial = 94; p = not significant)19

1 vs. 1 at one year (n = 18; p = not significant)14

2 vs. 7 at unspecified times (n = 30; p < 0.05)15

3 vs. 11 at a mean of 57 ± 24 days vs. 73 ± 59 days respectively (not pre-specified according to the

article (n = 70; p = 0.016)17

4 vs. 24 at unspecified times (n = 100; p < 0.05)16

If the results are restricted to those from the two double-blind trials with clearly specified follow-up times (at

one year, the first two trials listed), then it is clear that there were too few amputations to show any difference

in the incidence of major amputation. The authors of the first trial (involving 94 patients) suggested that there

may be differences in criteria used for amputation in the different studies, accounting for the large differences

in overall rates (e.g. note the incidences in the first trial, conducted in Sweden,19 and the last trial, conducted

in Turkey16). Furthermore, if a trial investigator is involved in the decision to amputate, there may be bias in

the trials that are not double-blind in design. The quality of standard treatment to treat the ulcer can also be a

confounding factor if it is different among the groups. Sufficient information was provided to deduce that such

treatment was optimal in three trials14,17,19 but not in the other three trials.15,16,18

Minor amputation. The incidence of minor amputation was reported in four trials but no indication of the time

from treatment was given. The differences were reported to be significantly different between the groups in

one trial (0 vs. 17)16 but not in the others.14,15,17

Chronic refractory radiation-induced proctitis

Evidence

One multicentre, double-blind, RCT randomised 150 patients with refractory “late” radiation-induced proctitis

to receive 30 sessions of either HBO at 2.0 ATA or normobaric air at 1.1 ATA, for 90 minutes per session,

over a period of 6 weeks.22 (The time from radiation to the diagnosis of late radiation-induced injury ranged

from 2.5 to 155 months in the patients included in the trial.) Measures were taken to preserve blinding of the

control group by a brief compression to 1.34 ATA followed by decompression to 1.1 ATA. After 30 sessions,

10 further sessions were provided to selected patients at the discretion of the referring physician, who

remained unaware of the allocation. Subsequently, unblinding took place, and the HBO group entered the

follow-up phase, with assessment occurring at 3 and 6 months, then at 1, 2, 3, 4, and 5 years. At the time of

unblinding, the patients in the control group were crossed over to the active treatment arm, but it is not clear

from the article for how long these patients were then treated; this time period would have extended into the

follow-up phase.

The SOMA-LENT score (a measure of the severity of radiation-induced complications with a maximum score

of 56) was 12.55 and 12.84 in the HBO and sham groups respectively, at the start of treatment. In both

groups, the score decreased (indicating improvement) over the six-week double-blind treatment phase, with a

significantly greater decrease in the HBO group compared with the sham group (decrease = 5.00 vs. 2.61

points, p = 0.015). After the sham group crossed over to active treatment, the difference between the two

groups became not significant, but it is not clear for how long treatment with HBO was continued or at what

time point this was measured. During follow up, improvement continued, with no significant differences being

reported for the scores in the two groups. At one year, when 105 patients remained in the study, the scores

were 5.29 and 6.72 respectively, and at two years (n = 61) they were 3.61 and 6.20 respectively (p values

were not reported for times after “completion of crossover”). For the secondary outcome involving a

standardised clinical assessment, the percentage of patients considered to be healed or improved was

significantly higher with HBO than with sham treatment (88.9% vs. 62.5% respectively) at the end of the

double-blind phase. The other secondary outcomes were related to quality of life issues; no statistically

significant differences were reported between the groups. 

Indications for which there was clearly inadequate evidence (in quantity or quality) to support use of HBO

The published trials relating to these indications, as described in the QIS Report and relevant Cochrane

reviews, was considered to be inadequate to support use of HBO because of absence of controlled trials,

conflicting results, poor quality in design (e.g. underpowered because of too few patients, non-clinical

surrogate outcomes, high risk of bias), or inadequate reporting (e.g. of randomisation procedure, duration of

follow-up, patient population not described adequately, insufficient details of methods used). These limitations

made it impossible to draw any conclusions regarding the efficacy of HBO.

On-going studies

Two on-going studies have been mentioned under CO poisoning and diabetic foot ulcer above. Other ongoing trials throughout the world that have clinical outcomes and are registered with ClinTrials.gov, the

ISRCTN Register and the NIHR, include the following:

Controlled trials

HBO therapy:

• as post-operative therapy to reduce complications in diabetic patients undergoing vascular surgery

(double-blind RCT, sham control)

• in calcaneal intra-articular fractures (open-label RCT, sham control)

• to prevent osteonecrosis in patients who have taken bisphosphonates (open-label RCT, untreated

control)

• for interstitial cystitis (open-label RCT, interstitial instillation of DMSO as control)

• for trigeminal neuralgia (double-blind RCT, sham control)

• to improve erectile function following surgery for prostate cancer (double-blind RCT, sham control)

• to treat mandibular osteoradionecrosis

• for prevention of osteoradionecrosis in patients requiring surgery to the mandible who have previously

undergone radiotherapy

Uncontrolled trials

HBO therapy:

• as adjuvant treatment for frost injury

• for persistent post-concussive symptoms after mild traumatic brain injury

• for mandibular osteoradionecrosis

• in patients with white matter hyperintensities

• for treatment of traumatic brain injury

• for post-concussive symptoms 

for the treatment of osteoradionecrosis

• as preconditioning in cardiovascular surgery

• for insulin resistance

• In children with autism (5 uncontrolled studies)

• for lower leg trauma

• for retinitis pigmentosa

• for trigeminal neuralgia

• in chronic traumatic brain injury or post-traumatic stress disorder

• with radiotherapy and temozolomide in adults with newly diagnosed glioblastoma

• to treat long-term gastro-intestinal adverse effects caused by radiation therapy in patients with pelvic

cancer

One study of acute CO poisoning in comatose patients was halted after the first interim analysis as there were

significantly more patients who recovered at one month in the control arm than in the HBO arm. Another

study of acute domestic CO poisoning was terminated “for futility” (no further details provided).