Chapter  85:  Hyperbaric Oxygen Therapy for Brain Injury, Cerebral Palsy, and Stroke

Technical Expert Advisory Group (TEAG)

We identified technical experts to assist us in formulating the research questions and identifying relevant databases for the literature search. The expert panelists included a neurologist specializing in stroke, a neurosurgeon specializing in severe brain injury, a pediatric neurologist with expertise in treating patients with cerebral palsy, and a physician with an HBOT practice. Throughout the project period, we consulted individual members of the TEAG on issues that arose in the course of identifying and reviewing the literature.

Literature Search, Study Selection, and Data Extraction

We searched a broad range of databases to identify published and unpublished studies of the effectiveness and harms of HBOT in patients with brain injury, cerebral palsy, and stroke. Each database was searched from its starting date to March 2001. The databases searched were:

• MEDLINE^®

• PreMEDLINE^®

• EMBASE

• HealthSTAR (Health Service Technology, Administration and Research)

• CINAHL^® (Cumulative Index to Nursing & Allied Health)

• Cochrane Database of Systematic Reviews

• Cochrane Controlled Trials Register

• DARE (Database of Abstracts of Reviews of Effectiveness)

• AltHealth Watch

• MANTIS™ (Manual, Alternative and Natural Therapy)

• Health Technology Assessment Database

TEAG members identified the following additional databases as potential sources of other material that may not be indexed in other electronic databases:

• The Undersea & Hyperbaric Medical Society: a large bibliographic database

• The Database of Randomised Controlled Trials In Hyperbaric Medicine

• European Underwater and Baromedical Society

• International Congress on Hyperbaric Medicine

• National Baromedical Services, Inc.

Update literature searching of the electronic databases MEDLINE, PreMEDLINE, EMBASE, CINAHL, the Cochrane Library, and the Health Technology Assessment Database was completed on February 26, 2002, using the same search strategy as used for the initial searches. Eight additional references submitted by a peer reviewer were added in May 2003. Finally, a supplemental search of MEDLINE, PreMEDLINE, EMBASE, and CINAHL was conducted in July 2003.

The references of all included papers were hand searched. In addition, two reviewers independently conducted hand searches of the references from the Textbook of Hyperbaric Medicine.¹ One TEAG member provided articles and meeting abstracts from his personal library.

Two reviewers independently assessed each title and abstract located through the literature searches for relevance to the review, based on the intervention, population, outcome, and study design criteria. The full-text articles, reports, or meeting abstracts that met the criteria listed above were retrieved and reviewed independently by two reviewers who reapplied the eligibility criteria. Disagreements were resolved through consensus.

Extraction of data from studies was performed by one reviewer and checked by a second reviewer. Disagreements were resolved through consensus.

Internal and External Validity and Quality Rating

The quality of all trials in the review was assessed using a list of items indicating components of internal validity. We modified the standard checklists to address issues of particular importance in studies of HBOT. For randomized controlled trials (RCTs) and nonrandomized controlled trials (NRCTs), the items assessed for internal validity were: randomization/allocation concealment, baseline comparability of groups, timing of baseline measures, intervention, outcome measures, timing of followup measurements (long enough to assess effects), loss to followup, handling of dropouts or missing data, masking, statistical analysis (if any), and general reviewer comments.

For the observational studies, items assessed for internal validity were exposure measurement (whether all subjects were given the same HBOT treatment), other interventions, differences in baseline factors among the groups of subjects compared (if a comparison group was included), discussion of or control for potential confounding, masking, evidence of stable baseline, timing of baseline survey, timing of followup measures, outcome measures used, and general comments of the reviewer.

Each study was then assigned an overall rating (good, fair or poor) according to the US Preventive Services Task Force method:

• Good: Comparable groups assembled initially (adequate randomization and concealment, and potential confounders distributed equally among groups) and maintained throughout the study; followup at least 80 percent; reliable and valid measurement instruments applied equally to the groups; outcome assessment masked; interventions defined clearly; all important outcomes considered; appropriate attention to confounders in analysis; for RCTs, intention-to-treat analysis.

• Fair: Generally comparable groups assembled initially (inadequate or unstated randomization and concealment methods) but some question remains whether some (although not major) differences occurred with followup; measurement instruments acceptable (although not the best) and generally applied equally; outcome assessment masked; some, but not all, important outcomes considered; appropriate attention to some, but not all, potential confounders; for RCTs, intention-to-treat analysis.

• Poor: Groups assembled initially not close to being comparable or not maintained throughout the study; measurement instruments unreliable or invalid or not applied equally among groups; outcome assessment not masked; key confounders given little or no attention; for RCTs, no intention-to-treat analysis.

For each study, the reviewer's assessment of external validity is given, including an assessment of the evidence that the study population reflects the underlying patient population (age-range, co-morbidities, co-interventions, etc.). External validity indicates the applicability of the results of the study to clinical practice. For example, if the study recruited a narrowly defined group of patients, the results may not be generalizable to a broader spectrum of patients. A study can have high internal validity but low external validity. There are no well-defined criteria for assessing external validity, and clinicians must assess the applicability of the results to the patient population for which the intervention is intended.

Brain Injury

• For traumatic brain injury, one randomized trial provided fair evidence that HBOT might reduce mortality or the duration of coma in severely injured TBI (traumatic brain injuries) patients. However, in this trial, HBOT also increased the chance of a poor functional outcome. A second fair quality randomized trial found no difference in mortality or morbidity overall, but a significant reduction in mortality in one subgroup. Therefore, they provide insufficient evidence to determine whether the benefits of HBOT outweigh the potential harms.

• The quality of the controlled trials was fair, meaning that deficiencies in the design add to uncertainty about the validity of results.

• Due to flaws in design or small size, the observational studies of HBOT in TBI do not establish a clear, consistent relationship between physiologic changes after HBOT sessions and measures of clinical improvement.

• The evidence for use of HBOT in other types of brain injury is inconclusive. No good- or fair-quality studies were found.

Cerebral Palsy

• There is insufficient evidence to determine whether the use of HBOT improves functional outcomes in children with cerebral palsy. The results of the only truly randomized trial were difficult to interpret because of the use of pressurized room air in the control group. As both groups improved, the benefit of pressurized air and of HBOT at 1.3 to 1.5 atm should both be examined in future studies.

• The only other controlled study compared HBOT treatments with 1.5 atm to delaying treatment for 6 months. As in the placebo-controlled study, significant improvements were seen, but there was not a significant difference between groups.

• Two fair-quality uncontrolled studies (one time-series, one before-after) found improvements in functional status comparable to the degree of improvement seen in both groups in the controlled trial.

• Although none of the studies adequately measured caregiver burden, study participants often noted meaningful reductions in caregiver burden as an outcome of treatment.

Stroke

• Although a large number of studies address HBOT for the treatment of stroke, the evidence is insufficient to determine whether HBOT reduces mortality in any subgroup of stroke patients because no controlled trial assessed was designed to assess mortality.

• Among controlled trials, the evidence about morbidity is conflicting. The three best-quality trials found no difference in neurological measures in patients treated with HBOT versus patients treated with pressurized room air.

• Two other controlled trials, one randomized and one nonrandomized, found that HBOT improved neurological outcomes on some measures. However, both were rated poor-quality.

• Most observational studies reported favorable, and sometimes dramatic, results, but failed to prove that these results can be attributed to HBOT. For example, one retrospective study found better mortality rates in patients who received HBOT than a comparison group of patients from a different hospital who did not. The study did not provide information on mortality rates from other causes in each hospital; this information would have made it easier to judge whether the improved survival was due to HBOT or to differences in overall quality of care at the HBOT hospital.

• The observational studies of HBOT provided insufficient evidence to establish a clear relationship between physiologic changes after HBOT sessions and measures of clinical improvement. Few studies established that patients were stable at baseline.

Adverse Events

• Evidence about the type, frequency, and severity of adverse events in actual practice is inadequate. Reporting of adverse effects was limited, and no study was designed specifically to assess adverse effects.

• The few data that are available from controlled trials and cohort studies of TBI suggest that the risk of seizure may be higher in patients with brain injuries treated with HBOT.

• No study of HBOT for brain injury, cerebral palsy, or stroke has been designed to identify the chronic neurologic complications.

• Pulmonary complications were relatively common in the trials of brain-injured patients. There are no reliable data on the incidence of aspiration in children treated for cerebral palsy with hyperbaric oxygen.

• Ear problems are a known potential adverse effect of HBOT. While ear problems were reported in brain injury, cerebral palsy, and stroke studies the incidence, severity and effect on outcome are not clear. However, the rates reported among cerebral palsy patients were higher (up to 47 percent experiencing a problem) than reported with brain injury or stroke. However, the data in brain injury are limited by the use of prophylactic myringotomies.

Supplemental Qualitative Analysis

• Opinions about the frequency and severity of risks of HBOT vary widely.

• Several participants emphasized the importance of continued treatments to maximize results.

• Patients and caregivers value any degree of benefit from HBOT highly. An improvement that may appear small on a standard measure of motor, language, or cognitive function can have a very large impact on caregiver burden and quality of life.

Outcome Studies

We identified several barriers to conducting controlled clinical trials of HBOT for brain injury, particularly cerebral palsy:

• Lack of agreement on the dosage and the duration of treatment.

• Need for better measures of relevant outcome measures, such as caregiver burden.

• Lack of independent, reliable data on the frequency and severity of adverse events.

• Patients' unwillingness to be assigned to a placebo or sham treatment group.

As described below, strategies can be developed to conduct good-quality studies to overcome each of these barriers.

Dose and duration of treatment. Oxygen, the “active ingredient” in HBOT, is fundamentally a drug. As for any drug, dose and duration of treatment must be determined in carefully designed dose-ranging studies before definitive studies demonstrating clinical efficacy can be started. Good-quality dose-ranging studies of HBOT for brain injury can be done, based on the model used by pharmaceutical manufacturers and the FDA. It is likely that the dosage of HBOT needs to be individualized based on the patient's age, clinical condition, and other factors. This is the case for many other drugs and does not pose an insurmountable barrier to designing dose-finding trials. In fact, the need to individualize therapy makes it essential to base the design of long-term studies of clinical outcomes on the results of dose-ranging studies.

Better outcome measures. In describing the course of their patients, experienced clinicians who use HBOT to treat patients with brain injury, cerebral palsy, and stroke refer to improvements that may be ignored in standardized measures of motor and neuro-cognitive dysfunction. These measures do not seem to capture the impact of the changes that clinicians and parents perceive. Caregivers' perceptions should be given more weight in evaluating the significance of objective improvements in a patient's function. Unfortunately, studies have not consistently measured caregiver burden, or have assessed it only by self-report. Studies in which the caregivers' burden was directly observed would provide much stronger evidence than is currently available about treatment outcome.

Adverse events. Uncertainty about the frequency and severity of serious adverse events underlies much of the controversy about HBOT. The case against HBOT is based on the reasoning that, because HBOT may be harmful, it must be held to the highest standard of proof. A corollary is that, if HBOT can be shown to be as safe as its supporters believe it to be, the standard of proof of its efficacy can be lowered.

Good-quality studies of adverse effects are designed to assess harms that may not be known or even suspected. The most common strategy is to use a standard template of several dozen potential adverse effects affecting each organ system. Other characteristics of a good study of adverse events are a clear description of patient selection factors, independent assessment of events by a neutral observer, and the use of measures for the severity (rather than just the occurrence) of each event.

Unwillingness to be in a placebo group. The issue of placebo groups has been the subject of a great deal of debate. Participants on both sides make the assumption that an “evidence-based” approach implies devotion to double blind, placebo-controlled trials without regard to practical or ethical considerations. This assumption is false. Double blind, placebo-controlled trials are the “gold standard” for government regulators overseeing the approval of new pharmaceuticals, but not for clinical decision making or for insurance coverage decisions. Evidence-based clinical decisions rely more heavily on comparisons of a treatment to other potentially effective therapies than to placebos.

Several alternatives to the double blind, placebo-controlled trial can be used to examine effectiveness. One approach is to compare immediate to delayed treatment with HBOT, as was done in the Cornell trial. Another is to design a trial in which patients are randomly assigned to several alternative HBOT regimens. Because of uncertainty about the dosage and duration of treatment, such a trial would be preferable to a trial that offered a choice between one particular regimen and no treatment at all. It is also easier to incorporate a sham therapy arm in such a trial: patients may be more willing to enter a trial if they have a 10 percent or 20 percent chance of being assigned to sham treatment instead of a 50 percent chance. Other alternatives to a placebo include conventional physical, occupational, and recreational therapy, or another alternative therapy, such as patterning.

The Canadian trial of HBOT for cerebral palsy has important implications for the design of future research. In the trial there was a clinically significant benefit in the control group. Debate about the trial centers largely on how the response in the control group should be interpreted. The trial investigators believe that the beneficial effect was the result of the psychological effect of participating in the trial and extra attention paid the children in and out of the hyperbaric chamber. Alternatively, the slightly pressurized air (that is, “mild” hyperbaric oxygen) may have caused the improvement. A third possibility is that the slightly increased oxygen concentration, not the pressure per se, was responsible for the benefit.

A trial that could sort out which of these explanations was true would have a major impact on clinical practice. Such a trial might compare (1) room air under slightly elevated pressure, delivered in a hyperbaric chamber, to (2) elevated oxygen concentration alone, delivered in a hyperbaric chamber, and to (3) an equal amount of time in a hyperbaric chamber, with room air at atmospheric pressure. From the perspective of a neutral observer, the third group is not a “sham” but rather an attempt to isolate the effect of the social and psychological intervention cited by the Canadian investigators.

In addition to needing improved design, future trials of HBOT need better reporting. This would aid interpretation and the application of the research results. Two types of information are essential: a clear description of the research design, particularly of the control and comparison groups, and a detailed description of the patient sample. It is frequently difficult to tell from published studies how comparable the patient populations are, not only demographically but also clinically, in order to interpret the diagnosis and prognosis.

Traumatic Brain Injury

Each year, approximately 1.5 million Americans sustain traumatic brain injuries, ranging in severity from mild to fatal.^{3, 4} The leading causes of traumatic brain injury are motor vehicle crashes, falls, firearm use, and sports and recreational activities. Adolescents and young adults (aged 15 to 24) as well as adults aged 65 years and older have the highest risk. The annual costs of traumatic brain injuries are estimated to be $56 billion.⁵ This figure reflects the costs of medical care and rehabilitation as well as the loss of productivity and income among individuals who have long-term disability due to their injuries.

Of the 1.5 million who are injured each year, 50,000 die, and from 80,000 to 90,000 experience the onset of long-term disability.³ The Centers for Disease Control has estimated that 5.3 million Americans are living with disability as a result of brain injury. The types of disability range across the entire spectrum of human physical, social, and emotional function. No single instrument can measure all of the consequences of TBI. The two most commonly used measures of outcome were developed for use with severe brain injury. The oldest formal scale, the Glasgow Outcome Scale (GOS), categorizes patients into five broad categories: good recovery, moderate disability, severe disability, persistent vegetative state, and death.⁶ This measure is convenient and very widely used, but it is insensitive to many of the cognitive and emotional deficits that, while subtle, have a strong effect on quality of life.

Another commonly used instrument is the Disability Rating Scale (DRS), a 30-point scale based on ratings of the level of consciousness or arousal, cognitive ability for self-care, physical dependence on others, and ability to work.⁷ Other measurement instruments have been designed to assess subtler degrees of disability in memory, cognition, attention, social function, or emotional function in chronic brain injury patients. These measures include the Community Integration Questionnaire, the Neurobehavioral Functioning Inventory (NFI), the Patient Competency Rating Scale (PCRS), the Level of Cognitive Functioning Scale (LCFS), and the Revised Craig Handicap Assessment and Reporting Technique (R-CHART).

Table 1. Glasgow Coma Scale

Table 1. Glasgow Coma Scale

Eye Opening Response

• Spontaneous--open with blinking at baseline 4 points

• To verbal stimuli, command, speech 3 points

• To pain only (not applied to face) 2 points

• No response 1 point

Verbal Response

• Oriented 5 points

• Confused conversation, but able to answer questions 4 points

• Inappropriate words 3 points

• Incomprehensible speech 2 points

• No response 1 point

Motor Response

• Obeys commands for movement 6 points

• Purposeful movement to painful stimulus 5 points

• Withdraws in response to pain 4 points

• Flexion in response to pain (decorticate posturing) 3 points

• Extension response in response to pain (decerebrate posturing) 2 points

• No response 1 point

Table 2. Results for Study by Artru, Chacornac, and (more...)

Table 2. Results for Study by Artru, Chacornac, and Deleuze

	HBOT Group	Control Group
Died within 1 year	15/31	16/29
Conscious at 1 month	13/31	8/31
Independent in daily activities at 1 year among survivors	14/31	12/29

GCS has several limitations as a predictor for an individual's outcome. Data about the ability of GCS scores to predict functional outcomes come from patients who undergo inpatient rehabilitation, rather than from all patients who are seen for trauma.¹⁴ Patients excluded from inpatient rehabilitation because they are not as severely impaired, are too impaired to benefit, or lack financial resources, have not been well studied. In trauma patients, mortality rates differ between different groups of patients who have similar average GCS scores. For a given GCS score, survival also varies considerably among published studies.¹⁵ The inter-observer reliability of the GCS is only fair, and the differences between observers are large enough to alter the predicted prognosis substantially.¹⁶ Patients who have the same total GCS score, but different scores on the components of the GCS, have different mortality rates.¹⁷

In addition to GCS, factors such as age,^{15, 18, 19} associated injuries,²⁰ intracranial hypertension,^{21, 22} and the presence of a mass effect²³ are also predictors of mortality and severe disability. Preinjury productivity and education also help predict functional outcome in survivors.^{24, 25} Hypoxia (defined as PaO2 less than 60 mm Hg, or apnea or cyanosis in the field) and hypotension (defined as a measure of systolic blood pressure less than 90 mm Hg at any time) are also strong predictors of death and severe disability.^{22, 26–28}

Some features of the patient's course and management in the hospital are also predictors of mortality and morbidity. For example, the extent of post-traumatic amnesia (PTA) is correlated with the prognosis. The longer the amnesia occurs following the injury, the worse the prognosis for recovery. If the loss of consciousness lasts more than 4 weeks, a high prevalence of impairment, inattention, and memory loss is predicted.²⁹

During the course of intensive care, episodes of hypotension, elevated intracranial pressure, decreased cerebral perfusion, and hypoxia are also predictors of a poor outcome.³⁰ Such episodes are very common. In a series of 184 patients receiving intensive care for acute, severe TBI,³⁰ all but seven patients had at least one episode of hypotension. In 157 of these patients, jugular venous oxygen saturation was monitored continuously. Ninety-seven (62 percent) of these patients experienced one or more episodes of hypoxia (jugular venous oxygen saturation <50 percent), and patients spent an average of 1.88 hours in a hypoxic state during the intensive care unit stay. These figures probably represent better-than-typical results because they were obtained in an intensive care unit that used invasive monitoring to minimize the frequency and duration of hypoxic episodes.

GCS and other prognostic factors are of little value in predicting the speed of recovery from coma. In before-after comparison studies, a presumption is often made that a patient who was discharged from the acute care hospital in a vegetative state has a very low chance of recovering consciousness spontaneously. However, several cases of recovery have been documented in patients who have had stable coma for longer than 6 months.³¹ In case series of patients with severe or catastrophic traumatic brain injuries, three of four patients who survived 6 months regained consciousness.¹⁸

Similarly, data from recent followup studies contradict the widely held view that improvements in neurocognitive function are unlikely to occur if more than a year has passed since the injury.^{32, 33} In one of these cohort studies, patients with TBI were administered a battery of 12 neuropsychological tests 1 year and 5 years after injury.³³ On one of the tests (Trails B, a test of complex attention), 22.2 percent of patients improved and 14.1 percent deteriorated between 1 and 5 years post-injury. For the other 11 tests, 0 to 22.7 percent (median 13 percent) improved and 1.2 percent to 18 percent (median 6.2 percent) deteriorated. The authors concluded that clinically significant improvements can occur long after apparently “stable” deficits have been diagnosed.

Nearly 90 percent of TBIs that are reported annually in the United States are classified as mild TBI (MTBI) or concussion. MTBI can cause immediate neurocognitive abnormalities³⁴ as well as long-term problems such as persistent headaches, confusion, memory problems, mood changes, and changes in vision or hearing.³⁵ The incidence of MTBI may be under-reported because a large percentage of patients never seek medical evaluation or treatment. Moreover, the subtle long-term consequences of MTBI, although often apparent to patients and family members, may go unrecognized by physicians. It is difficult to predict which patients with MTBI will suffer long-term disability.

Many components of acute and rehabilitative care for patients with brain injuries are not supported by good-quality evidence from clinical trials.^{27, 36, 37} A recent consensus conference on clinical trials in head injury summarized the disappointing results of over a dozen treatments which despite promising results in observational studies, proved to be ineffective when tested in randomized trials.³⁸ Most of these treatments appeared to be effective in animal studies, case reports, and other before-after studies.

Cerebral perfusion pressure (CPP) management provides the best example. Following several case reports and small series, Rosner and colleagues published a series of 158 patients admitted with a Glasgow Coma Scale score less than 7 who underwent Cerebral perfusion pressure (CPP) management rather than the conventional approach, control of intracranial pressure (ICP).³⁹ Mortality was only 29 percent, and 59 percent achieved a good recovery or moderate disability by 6 months post-injury. The authors stated that these mortality and recovery rates were much better than would be expected from other series of patients who had similar GCS scores. These results led to wide use of the CPP management strategy. However, in a subsequent randomized controlled trial that recruited patients with a GCS score less than 5, mortality was under 30 percent in both the cerebral blood flow-targeted and conventionally managed groups, and there was no difference in neurologic outcomes.⁴⁰ In this trial, the cerebral blood-flow-targeted strategy significantly reduced the frequency of cerebral ischemia and of jugular desaturation, but these physiologic improvements did not translate into clinical benefits.

Hypothermia provides a similar example. In the 1990s, several groups of investigators published dramatic case studies and series of cases that appeared to show that inducing hypothermia in brain-injured patients improved outcomes.^41–43 The goal of this treatment was to reduce hypoxia in the injured brain tissue. In one series, the investigators used hypothermia in 148 patients who had initial GCS scores less than 6.⁴¹ Mortality was 30 percent, and, as measured by the Glasgow Outcome Scale, 40 percent of patients had a good recovery, 13 percent had mild disability, and 10 percent were in a persistent vegetative state. Subsequently, in a randomized trial of 392 patients, mortality was 28 percent in the hypothermia group and 27 percent in the normothermia group. An additional 30 percent of patients in each group had severe disability or a vegetative state. The hypothermia group had fewer episodes of high intracranial pressure but also had more hospital days because of complications.

These examples show that improvements in physiologic measures do not always translate into tangible clinical results. They also show that relying on assumptions about the expected prognosis of a group of brain-injured patients, rather than on results in a control group, can be misleading.¹ Even over a short time there can be significant changes in the prognosis of TBI. In one trauma center, for example, mortality among patients with a GCS ≥ 4 fell from 40 percent in the period 1980-1981 to 27 percent in 1987-1988 and 2.8 percent in 1996-1997.⁴⁴

Anoxic-ischemic Brain Injury

It is estimated that, in the U.S., more than 1,000 useful lives are lost each day as a result of poor cardiopulmonary and trauma resuscitation outcomes.⁴⁵ Among those who survive, permanent brain injury is a common, devastating complication. In addition to cardiopulmonary arrest, toxic substances, congenital disorders, and birth trauma can cause brain injury by means of anoxia and ischemia.

Prediction of the outcome of coma due to anoxic-ischemic coma is poor. In a meta-analysis of studies of patients with anoxic-ischemic coma, the sensitivity of a GCS score of 3 to 5 ranged from 63 percent to 95 percent for a poor outcome, defined as death or persistent vegetative state.⁴⁶ The specificity of a GCS in this range was 54 percent to 100 percent. The meta-analysis found that clinical variables are less accurate in predicting outcome after 24-hour coma duration than after 72 hours of coma. The most specific predictors of outcome were the lack of pupillary light reflexes after 72 hours, lack of motor response to pain after 72 hours, and certain somatosensory evoked potential findings. A subsequent meta-analysis by the same authors found there was insufficient evidence to determine whether markers of central nervous system metabolism added substantially to the predictive value of these variables.⁴⁷

Cerebral Palsy

Each year, about 10,000 babies born in the U.S. develop cerebral palsy. More than 500,000 Americans have cerebral palsy. A study in California showed that the lifetime costs per new case of cerebral palsy was $503,000 (in 1992 dollars).⁴⁸ Half of these costs are borne by families, who often find it difficult to obtain all the services they need to help their children.

Cerebral palsy results from injury to the brain. About 20 percent of children who have cerebral palsy acquire the disorder after birth, while 80 percent of cases are congenital. Meningitis, encephalitis, and trauma cause most of the acquired cases. According to the National Institute of Neurological Disorders and Stroke, the mechanism of injury in the majority of cases of congenital cerebral palsy is not known. Until recently, the belief that birth complications cause most cases of cerebral palsy was widespread. Then, in the 1980s, a careful study of 35,000 births showed that fewer than 10 percent of children with cerebral palsy had a history of birth complications such as rubella or other infections during pregnancy, jaundice, Rh incompatibility, asphyxia (oxygen shortage), or head trauma during labor and delivery. Most children with congenital cerebral palsy do not have a history of any of these conditions. Premature birth and low birthweight predispose to cerebral palsy, but the reason for this association is not clear.

Cerebral palsy represents a very broad range of motor disorders, varying in the part of the body they affect (e.g., diplegia, hemiplegia, quadriplegia); the type of motor disorder (spastic, athetoid, or ataxic) and their severity. The most familiar pattern is spastic diplegia, meaning that the patient has stiff, contracted muscles in the legs. By definition, the muscle disorder in cerebral palsy is not progressive. However, muscle spasticity, even if stable, can cause new problems as a child grows. For example, pain and contractures may increase as the bones of the child's legs lengthen.

Standardized scales, such as gait analysis, and functional scales, such as the Gross Motor Function Measure (GMFM), are used to assess and monitor progress. The GMFM is a validated and reliable scale used for measuring function in patients with cerebral palsy. It consists of five domains with a possible total score of 88. Various prognostic criteria for the patient's function have been developed over the years. For example, if a patient is not sitting independently when placed by age 2, then one can predict with approximately 95 percent confidence that he/she never will be able to walk.⁴⁹ On occasion, such a child will walk, but usually aids are necessary, such as a walker. Most children with cerebral palsy will improve in their function over time, ⁵⁰ but many have deficits that continue into adulthood.

Stroke

Mortality and morbidity from a stroke are related to older age, history of myocardial infarction, cardiac arrhythmias, diabetes mellitus, and the number of stroke deficits.⁵¹ Evaluation by magnetic resonance imaging (MRI) of the brain obtained during the first few days of the stroke will predict a favorable outcome if less than 80 cc of the brain is infarcted.⁵² The 30-day survival after a first stroke has been estimated to be less than 80 percent.⁵³ For those who survive, it has been estimated that 95 percent of patients reach maximal recovery within 3 to 5 months of the stroke.⁵⁴ Functional recovery is dependent on numerous variables, including age, neurologic deficit, comorbidities, psychosocial factors, educational level, vocational status, and characteristics of the stroke survivor's environment.

Hyperbaric Oxygen Therapy

Hyperbaric oxygen therapy is the inhalation of 100 percent oxygen inside a hyperbaric chamber pressurized to greater than 1 atmosphere (atm). HBOT causes both mechanical and physiologic effects by inducing a state of increased pressure and hyperoxia. Hyperbaric oxygen pressure is expressed in multiples of atmospheric pressure at sea level, where 1 atm is about 760 mm Hg or 1 kilogram per square centimeter.^{55, 56} The oxygen dissolved in blood at 1 atm (sea level) breathing room air is 0.3 ml/dL, and this is in addition to hemoglobin-bound oxygen. Breathing 100 percent oxygen at 1 atm results in an increase in blood oxygenation to 1.5 ml/dL. Increasing the pressure to 3 atm increases the blood oxygen (dissolved oxygen, not carried by hemoglobin) to 6 ml/dL.^{57, 58} At rest and with good perfusion, tissues require 5–6 ml/dL of oxygen, whether from dissolved or hemoglobin-bound oxygen. Hence, in situations where hemoglobin-bound oxygen is limited (e.g., carbon monoxide poisoning), tissue oxygen needs can be met without hemoglobin-carried oxygen.

In addition to this hyperoxic effect, the increased pressure reduces the volume of gases in the blood by virtue of Boyle's law (in an enclosed space, the volume of a gas is inversely proportionate to the pressure exerted upon it). This is the mechanism relied upon in decompression illness and arterial gas embolism to reduce the size of the gas bubbles and allow replacement of inert gas in the bubbles with oxygen, which can be metabolized by tissues.

HBOT can be administered in two primary ways, using a monoplace chamber or a multiplace chamber.^{59, 60} The monoplace chamber serves one patient at a time. It is the less-costly option for initial setup and operation but provides less opportunity for patient intervention while in the chamber. Monoplace chambers are generally constructed of clear acrylic or with acrylic view ports that allow for patient observation. Monoplace chambers are generally pressurized with 100 percent oxygen.

Multiplace chambers allow medical personnel to work in the chamber and care for acute patients to some extent. Each patient is given 100 percent oxygen through a facemask, tight-fitting hood, or endotracheal tube. The entire multiplace chamber is pressurized with air, so medical personnel may require a controlled decompression, depending on how long they are exposed to the hyperbaric air environment.

While the duration of an HBOT session is typically 90 to 120 minutes, the duration, frequency, and cumulative number of sessions has not been standardized. The dose received by the patient may be affected by the type of chamber used. Monoplace chambers using face masks or hoods that do not fit snugly may result in dilution of 100 percent oxygen with room air.⁵⁹

Intervention

• Hyperbaric Oxygen Therapy: any treatment using 100 percent oxygen supplied to a patient inside a hyperbaric chamber that is pressurized to greater than 1 atm; any frequency, duration, and total number of treatments.

Population

• Patients with brain injury from any cause and in any stage (acute, subacute, or chronic).

• Patients with cerebral palsy of any etiology.

• Patients with thrombotic stroke, excluding patients with transient ischemic attack (TIA), hemorrhage (e.g., subarachnoid hemorrhage), or vasospasm.

• Patients with progressive neurologic diseases (i.e., multiple sclerosis, Parkinson's disease, Alzheimer's disease, and chronic cerebral insufficiency), acute infectious processes (i.e., mucormycosis), radiation sensitization of brain tumors, and reports of treating eye damage or sudden deafness were excluded.

• The use of HBOT for approved indications such as acute carbon monoxide poisoning or acute air embolism was also excluded.

Outcomes

We sought articles reporting any clinical endpoint. In general we excluded studies that reported only intermediate outcomes, such as changes in cerebral metabolism or EEG findings. However, we included studies that reported the effect of HBOT on elevated intracranial pressure, an intermediate outcome that is currently a main determinant of treatment in current clinical practice.

Design

• We included studies of human subjects that reported original data (no reviews of studies).

• 

  

  Figure 1. HBOT Literature search results

  

     Figure 1. HBOT Literature search results

  TBI = traumatic brain injury; BI = brain injury

  We used the algorithm in Figure 1 to classify the design of studies. All of the study designs in the figure were included in the review except for non-comparative studies (e.g., case reports).

• Before-after or time-series studies with no control group were included if (a) five or more cases were reported, and (b) outcome measures were reported for both the pre- and post-HBOT period.

Electronic Database Literature Search

We searched a broad range of databases to identify published and unpublished studies of the effectiveness and harms of HBOT in patients with brain injury, cerebral palsy, and stroke. Each database initially was searched from its starting date to March 2001. Full details of all the strategies, the databases searched, the inclusive dates searched, the software used to search, and the number of citations found that were used in this review are provided in Appendix D.

The databases we searched were:

• MEDLINE

• PreMEDLINE

• EMBASE

• HealthSTAR (Health Service Technology, Administration and Research)

• CINAHL (Cumulative Index to Nursing & Allied Health)

• Cochrane Database of Systematic Reviews

• Cochrane Controlled Trials Register

• DARE (Database of Abstracts of Reviews of Effectiveness)

• AltHealthWatch

• MANTIS (Manual, Alternative and Natural Therapy)

• Health Technology Assessment Database

If only studies found in the large electronic databases are included, a publication bias may arise in the review. Studies with a positive and statistically significant finding are more likely to be published than those finding no difference between the study groups.⁸⁴ Because small studies are more likely to have negative results, this bias has also been called “sample size bias.”

Excluding “gray literature” is another potential source of bias. The term “gray literature” refers to reports of studies that are difficult to find, largely because they either are unpublished or are published in sources that are not indexed by the large electronic databases. The Interagency Gray Literature Working Group described gray literature as “foreign or domestic open source material that usually is available through specialized channels and may not enter normal channels or systems of publication, distribution, bibliographic control, or acquisition by booksellers or subscription agents.”⁸⁵ Studies found in the gray literature are not inherently lower quality than those identified through electronic methods, although they are more likely to be small and to have inadequate power to show a difference if one exists.

To avoid publication bias, we asked TEAG members to identify additional databases as potential sources of other material, particularly gray literature, meeting abstracts, and conference proceedings, that may not be indexed in other electronic databases such as MEDLINE. They identified the following sources:

• The Undersea & Hyperbaric Medical Society: a large bibliographic database (30,000 records), http://www.uhms.org/library.htm

• The Database of Randomised Controlled Trials In Hyperbaric Medicine, http://hboevidence.com/

• European Underwater and Baromedical Society, http://www.eubs.org/

• International Congress on Hyperbaric Medicine, http://www.ichm.net/

• National Baromedical Services, Inc.

Each organization was contacted regarding searching their databases. A search of the Undersea & Hyperbaric Medical Society database was conducted by its librarian using our search strategy. A search of the Database of Randomised Controlled Trials in Hyperbaric Medicine was conducted online by the principal investigator. A TEAG member provided the proceedings for 11 of the 12 International Congress on Hyperbaric Medicine conferences (Proceeding number 1 is no longer available). National Baromedical Services, Inc., conducted a search of its database and sent a list of titles at the request of one of the TEAG members. The European Underwater and Baromedical Society did not respond to our requests for access to its database.

Hand Searches

The references of all papers were hand searched. In addition, two reviewers independently conducted hand searches of the references from the Textbook of Hyperbaric Medicine.⁶⁰ One TEAG member provided articles and meeting abstracts from his personal library. These submitted articles and abstracts were also independently assessed for inclusion by two reviewers.

Update Searches

Update literature searching of the electronic databases MEDLINE, PreMEDLINE, EMBASE, CINAHL, the Cochrane Library, and the Health Technology Assessment Database was completed on February 26, 2002, using the same search strategy as used for the initial searches. The results of these searches are summarized in Appendix D. In May 2003, we added eight additional publications brought to our attention by a peer reviewer. Finally, a supplemental search of MEDLINE, PreMEDLINE, EMBASE, and CINAHL was conducted in July 2003.

Traumatic Brain Injury

Controlled Trials

Mortality and morbidity results. The best evidence of the effect of HBOT on mortality and morbidity in TBI comes from two fair-quality randomized controlled trials (see Evidence Table 1). The first trial found that HBOT had no effect on mortality at 12 months.⁸⁸ In the more recent trial, there was a dramatic decrease in mortality 12 months after treatment, but HBOT did not improve the rate of a favorable functional outcome.^89–91

Artru, Chacornac, and Deleuze (1976) ⁸⁸ studied 60 patients with coma due to head injuries. These patients were stratified into nine subgroups based on the severity of coma and the presence of mass lesions and then were randomized to HBOT or to standard therapy. The stratification resulted in groups that were similar in terms of type of injuries, but the authors did not report whether the two resulting groups were similar in other important prognostic variables. After 12 months of followup, overall mortality was similar in both groups. The rate of recovery of consciousness at 2 weeks and 1 month was higher in the HBOT groups (42 percent vs. 28 percent), but this finding was not statistically significant.⁸⁸ The mean duration of coma was also shorter in the HBOT group but was not statistically significantly different (28.2 days vs. 32.7 days, p = NS). In one of nine subgroups, patients under 30 years old with brain stem contusion were more likely to recover consciousness by 1 month if they received HBOT. There were nine patients in each group; one died in each group, and there were six conscious in the HBOT group and one conscious in the control group at 1 month (p < 0.03).

Rockswold, Ford, et al. (1994) ^89–91 enrolled 168 of 272 (62 percent) potentially eligible patients with acute closed-head trauma. Of the 272 potentially eligible patients, 18 percent died within 6 hours of admission, 8 percent had contraindications to HBOT, 6 percent were not identified as potential subjects in time for randomization within 6 hours of admission, and consent could not be obtained for 6 percent (no details given on the baseline characteristics or outcome of these patients).

The 168 patients who were randomized had Glasgow Coma Scale Scores of 9 or less, 6 to 24 hours after admission with a severe head injury, or 6 to 24 hours after deterioration following admission for what appeared to be a mild or moderate injury.^89–91 This study did not describe the methods used to randomize patients. There were several differences between the HBOT and control groups at the start of the study. For example, more patients in the control group had an operative mass lesion (39 percent vs. 49 percent), while more patients in the HBOT group had intracranial pressures above 20 mm Hg (52 percent vs. 46 percent). Overall, the differences in prognostic variables did not seem to favor either the HBOT group or the control group. The authors did not report whether patients enrolled after deterioration were distributed evenly (these patients may have a worse prognosis, the results of the trial could be biased if they were not distributed equally in the two groups).

Table 3. Results of Study by Rockswold, Ford, et al (more...)

Table 3. Results of Study by Rockswold, Ford, et al

	HBOT Group	Control Group
Died within 1 year	14/84	26/84
Dead or severely disabled at 1 year	40/84	40/84

The main results of the trial are summarized in Table 3. After 1 year, patients who were assigned to HBOT treatment had lower mortality (17 percent vs. 31 percent), but there was no difference in the proportion of patients who were either dead or severely disabled. Additional analysis showed that HBOT reduced mortality in patients who had a GCS score of 4 to 6 or ICP > 20 mm Hg, but not in other subgroups of patients.

Differences between these two studies might explain the discrepant mortality results. First, Artru was a much smaller study and could have missed an important difference in mortality. Second, the HBOT protocols differed among these studies. The Rockswold trial used 1.5 atm, while the Artru study used 2.5 atm.⁸⁸ In the Rockswold trial, patients were treated for 60 minutes every 8 hours for 2 weeks or until the patient regained consciousness or died. In the Artru trial, treatments were given daily for 10 days,⁸⁸ followed by 4 days without treatment, followed by 10 days of treatment until the patient regained consciousness or died. Third, in most cases Rockswold began treatment within 24 hours of injury, while in Artru there was an average delay of 4.5 days between the onset of coma and the start of HBOT.

It is important to note that the control patients in Artru et al. had higher mortality (about 50 percent) than the control patients in Rockswold (34 percent). This could be due to differences in baseline prognosis or to differences in the standard treatments given to control patients. Unfortunately, it is difficult to compare the prognoses of the patients in the two studies, because the two studies used different scales to assess the prognosis of the head injury and provided different information about associated injuries and other comorbidity. Rockswold used the Glasgow Coma Scale, while Artru used a modified Jouvet scale. The two scales are said to be poorly correlated.⁹²

It is also difficult to compare the standard treatments given to patients in the control group. Because of the spread in the years of publication (1976 to 1994) and the different countries involved, it can be assumed that these treatments may have varied considerably. For example, in the more recent Rockswold trial, all patients had invasive ICP monitoring and phenytoin. In the Artru study, most patients in both groups received furosemide and mannitol, but these decisions were based on clinical judgments about the likelihood of elevated ICP.

Differences in study quality are unlikely to explain the discrepant results of the trials, but the limitations of both trials make the validity of their results uncertain. Of the two trials, Rockswold had better internal validity (that is, quality), because it masked outcome assessors and reported how the patients were selected from the potential pool of eligible patients and how many refused enrollment.⁹⁰ Neither study described the methods used to randomize patients. Some methods of randomization cannot prevent investigators from (knowingly or unknowingly) placing patients with a more favorable prognosis into the treatment group. Empirical evidence has shown that studies that do not describe the method of randomization report exaggerated effects.⁹³ This flaw is particularly important in studies of comatose trauma patients, because an experienced clinician can predict prognosis within groups of patients who have a similar severity-of-illness score.

Effect of HBOT on physiologic measures. In the Rockswold trial, nearly all of the observed reduction in mortality occurred in the subgroup of patients who had ICP ≥ 20 mm Hg prior to treatment.^89–91 One important question is whether this benefit corresponded to a reduction in ICP in these subjects. The measurements were taken every 15 minutes during HBOT and then hourly until the next treatment, and hourly in the control group. The mean peak ICP values in controls (no HBOT) and HBOT-treated patients were not significantly different. However, mean peak ICP was significantly lower for patients who received HBOT and myringotomy (n = 42) versus patients who received only HBOT (n = 37) and versus controls (22.2 mm Hg in HBOT plus myringotomy vs. 33.0 in HBOT alone and 30.3 in controls, p < 0.05). The authors theorized that the pain caused by increased otic pressure contributed to the maintenance of elevated ICP, and that the effect of HBOT can be seen once prophylactic myringotomies are performed. The time to the mean peak ICP in the two groups was not reported. The duration of effect was not clear. Comparisons to other specific treatments for elevated ICP given to control patients were not reported. The number of subjects with HBOT plus myringotomy, reported in Table 3 in the paper, differs from the number reported in the text.

Other physiologic parameters of the effect of HBOT in traumatic brain injury include cerebral blood flow, the arteriovenous oxygen difference, the cerebral metabolic rate of oxygen, and the distribution of cerebral blood flow as visualized by a SPECT scan. The correlation of clinical outcomes with these measures has not been examined in controlled trials of HBOT for traumatic brain injury.

Other Nontraumatic Brain Injury

Synthesis

Table 4. Controlled trials of HBOT in brain injury

Table 4. Controlled trials of HBOT in brain injury

Study, population, n	HBOT Protocol; Control	Results	Quality Rating
Traumatic brain injury
Artru, 197688	2.5 atm × 60 minutes × 10 days of treatment alternating with 4 days off until patient regained consciousness or died.	No significant difference on any comparison (persistent coma at 1 month, consciousness recovery rate at 1 month, death rate at 1 month, death rate at 1 year, mean duration of coma, except in 1 of 9 subgropus: patients less than age 30, not reacting in an adapted manner to painful stimuli, and not operated on)	Fair
France; Patients with head injuries and in a coma;	Control: Standard treatment	n=9
31 HBOT, 29 control		HBOT, 9 control.
		Persistent coma at 1 month: 22% (HBOT); 78% (control); p<0.03
		Independent in daily activities at 1 year among survivors: 45% (HBOT); 41% (control)
Rockswold, 1992,90 199491	1.5 atm × 60 minutes every 8 hours × 2 weeks or until the pt was brain dead or could consistently follow simple commands. Average 21 treatments per patient.	Mortality at 12 months: 14/84 (17%) (HBOT); 26/82 (32%) (control)	Fair
Minnesota; severe head injury at one institution, admitted 1983-1989.	Control: standard intensive neurosurgical care, all patients received phenytoin.	p = 0.04
Total GCS score of 9 or less for at least 6 hours; 84 HBOT, 84 control		Favorable outcome at 12 months (GCS score of 1 or 2) 44/84 (52%) (HBOT); 44/82 54%) (control)
		p = 0.99
Other brain injury
Jianhua, 1995106	1.8–2.0 atm × 90 minutes once daily × 10 days.	HBOT group: 18/47 (38%) curative, 25/47 (53%) effective, 3/47 (6%) ineffective (1 missing?).	Poor
China; children (aged 1 to 11 years) with viral cerebritis;	Control: supportive therapy, including drugs naofukang, naofuxin.	Control group: 8/45 (18%) curative, 20/45 (44%) effective, 17/45 (38%) ineffective.
47 HBOT, 45 control		HBOT vs control: % curative p < 0.05, % effective p > 0.05, % ineffective p < 0.001.

GCS=Glasgow Coma Scale; GOS=Glasgow Outcomes Scale; atm=atmospheres ; NS=non-significant; ICP=intracranial pressure; CT=comuterized tomography

Evidence about the effectiveness of HBOT in traumatic brain injury is conflicting (see Table 4). One trial found that HBOT reduced mortality after 1 year of followup, but survivors were much more likely to be completely or severely disabled than survivors in the control group.^89–91 The other trial found no difference in mortality after 1 year of followup.⁸⁸ There are many possible explanations for the discrepant results, including the size of the trials, the protocols used to deliver HBOT, the baseline condition of the subjects, and differences in management other than HBOT.

The quality of the controlled trials was only fair, meaning that deficiencies in the design add to uncertainty about the validity of results.⁹³ Neither trial of HBOT for TBI described the methods used to conceal randomization, and neither resulted in clearly similar baseline groups.

In a fair-quality observational study,⁹⁷ severely brain-injured patients had better aerobic metabolism for up to six hours after an HBOT treatment. This study did not attempt to link this physiologic improvement after HBOT sessions to measures of clinical improvement. Other observational studies reported clinical endpoints, but they used subjective methods to assess recovery and provided insufficient information to determine whether the outcomes attributed to HBOT would have been expected from the severity of injury and other prognostic characteristics of subjects.

Controlled Trials

Collet, Vanasse, et al. (2001) randomized children aged 3 to 12 years with cerebral palsy to a course of treatments with HBOT of 100 percent oxygen pressurized to 1.75 atm or a similar course of treatments with room air pressurized to 1.3 atm.^{119, 120} The control treatment of 1.3 atm of room air provides approximately the same alveolar partial pressure of oxygen as 100 percent oxygen without pressurization (1 atm) given by face mask. The children were a well-defined group, with perinatal anoxia and documented cerebral palsy, motor developmental age 6 months to 4 years, and psychological development 24 months or more. Children with prenatal or antenatal causes of cerebral palsy were excluded. No concurrent interventions were allowed, and other treatments were stopped prior to the study. The primary measure of outcome was the GMFM scale, a validated measure of motor function. Eight other outcome measures were also used. All assessments were done by masked physical therapists.

Motor function improved in both groups of children. At the conclusion of 40 treatments, the mean changes in GMFM were 2.9 in the HBOT group and 3.0 in the control group; the difference was not statistically significant. Six months after initiating treatment, the mean changes were 3.4 and 3.1, respectively. This represents an increase of approximately 6 percent, which is considered to be a meaningful improvement for a short period of time, compared to approximately 7 percent improvement on GMFM at 12 months with dorsal rhizotomy.

Cognitive outcomes were measured using neuropsychological tests, consisting of five tests with multiple components and parental assessments. Visual working memory, auditory attention, and self-control were improved in both groups. Speed of information processing, verbal working memory, and visual attention remained the same over the course of the study. The tests where improvements were seen were thought to be susceptible to a learning effect, meaning that performance may have improved with repetition of the tasks and testing procedures.

No significant differences were found on any of these outcome measures assessed at any time point, with two exceptions. When the caregivers' viewpoint was assessed with the Pediatric Evaluation of Disabilities Inventory (PEDI), the control group had significantly better mobility and social functioning. The actual data for these comparisons were not presented.

This trial was assigned a fair-quality rating. The strengths of the trial are central randomization, masked outcome assessment, and use of objective, validated outcome measures. The areas of potential concern are the allocation concealment method (sealed envelopes, which can potentially reveal the allocation), differences in groups at baseline in presumed cause and type of cerebral palsy, and a baseline difference of 9 points on the 88-point GMFM scale between the treatment and control groups.

It is unclear whether the children included in this trial are representative of children with cerebral palsy. No information is given on how the 196 children screened for inclusion in the study were identified. Of these 196 children, 43 percent of children screened were not enrolled, and 32 percent refused to participate. The baseline characteristics or GMFM scores for these children were not reported.

The Cornell Study (Packard 2000) was a trial in which two groups of children received HBOT, but one received treatments immediately after enrollment (n=12) and the other 6 months after enrollment (n=14).¹¹⁸ The trial has not been published in a peer-reviewed journal. The children were aged 1 to 5 years, and had “moderate to severe cerebral palsy” and developmental delay of at least 33 percent in one area (areas not defined). A variety of measures (Bayley II, Preschool Language Scale [PLS], Peabody Motor Scales, PEDI) were assessed by masked physical therapists or child psychologists at baseline, 1 month, 2 months, and 5 months. However, diaries kept by unmasked parents appear to be the primary outcome measure in this study.

After 6 months, parental diaries indicated 22 percent of the subjects had major gains in skills and 44 percent of children with visual impairments (four of nine) reported improvement, but the report did not say how many of the improved children had received HBOT. Significant improvements in the PEDI score were seen initially but dissipated by 6 months. Other masked assessments and the Bayley II, PLS, and Peabody scales showed no difference between the groups.

Because a full report is not available, we could not fully assess the quality of the trial. We assigned the study a preliminary rating of poor-quality, but this rating could improve when more information about the study design becomes available. The preliminary report lacks important details regarding randomization and allocation concealment methods, baseline comparability data, and any description of the methods used to analyze results. The external validity cannot be assessed until more information about the selection and baseline characteristics of the patients becomes available.

Observational Studies

None of these publications included an adequate description of how patients reported were selected or of the diagnostic criteria used to determine eligibility of subjects. None attempted to control for potential confounders. All used a combination of objective and subjective outcome measures, but none masked outcome assessments by using an independent rater. Finally, none of these studies described the scales used to rate outcomes in sufficient detail to assess their validity or reliability.

Figure 2. Study Design Algorithm*

   Figure 2. Study Design Algorithm*

* Zaza S, Wright-De Aguero LK, Briss PA, et al. Data collection instrument and procedure for systematic reviews in the guide to community preventive services. Am J Prev Med 2000;18(1 Suppl):44–74.

Chavdarov (2002) is a poor-quality before-after study of 50 children that reported improvements of 13 percent for motor, 6 percent for mental, and 7 percent for speech abilities 2 days after HBOT.¹²¹ Data for each scale used were not presented.

Machado (1989) is a retrospective study¹²² of 230 patients who received HBOT for cerebral palsy. Immediately after HBOT, 218/230 (95 percent) children had reduced spasticity, based on a rating scale, but actual data were not reported. In 82 of these children followed for 6 or more months (the others were lost to followup), 62/82 (76 percent) had persisting reduction of spasticity and better motor control (data not reported). Parents reported other types of improvement, such as better balance, more attentive, and more “intelligent,” with a reduced frequency of convulsions and episodes of bronchitis. Vague inclusion criteria, outcome measures, and timing of measurements make the results unreliable.¹²² This study reported reduction in spasticity based on a scale (1 to 100) developed by the authors; however, it is stated that this scale was developed over time and could not have been used on all patients reported. It is not clear which patients were assessed using this scale, and the data for those who were assessed were not reported. This study was rated poor-quality.

Abstracts and Conference Proceedings

Synthesis

There is insufficient evidence to determine whether the use of HBOT improves functional outcomes in children with cerebral palsy. Observational studies (dose ranging from 1.5 to 1.75 atm) reported improvements on subjective measures and on motor function as measured by the GMFM. In the two controlled trials, however, similar improvements were seen in children who did not receive HBOT, indicating that HBOT may not be the cause of improvements seen in the observational studies.

The best evidence comes from a fair-quality randomized controlled trial, which found that HBOT at 1.75 atm and 1.3 atm of room air had a similar effect on motor function. Improvements in the GMFM over 6 months were 5 to 6 percent in both groups, which is considered significant improvement for a short period of time, and which may be compared with approximately 7 percent improvement on GMFM at 12 months with dorsal rhizotomy.

Different explanations have been offered to explain the improvement in the children who were treated with pressurized room air. The authors of the trial thought that the children in both groups improved because participation in the study provided an opportunity for more stimulating interaction with their parents. This is speculative, however, because there was no evidence to suggest that the parents and their children had less time together, or less stimulating interaction, before the study began.

Another possible explanation is that the “sham” intervention—pressurized room air—was beneficial. The trial was designed to test the efficacy of oxygen, the “active ingredient” in HBOT and in room air. At 1.3 atm, pressurized room air provides a similar amount of oxygen as unpressurized 100 percent oxygen by mask. The possibility that pressurized room air had a beneficial effect on motor function should be considered the leading explanation.

Controlled Trials

Table 6. Controlled trials of HBOT in stroke

Table 6. Controlled trials of HBOT in stroke

Study, population, n	HBO Protocol; Control	Results	Quality Rating
Anderson, 1991127	100% oxygen at 1.5 atm × 1 hour every 8 hours × 15 treatments. Vitamin E 400 units given with each HBOT.	Average graded neurological exam scores	Fair
Minnesota; patients with ischemic cerebral infarction; 20 HBOT, 19 control.	Control: Room air at 1.5 atm × 1 hour every 8 hours × 15 treatments.	Baseline 44.6 (control), 45.5 (HBOT):
	Both groups received standard ICU plus physical and occupational therapy.	Day 5: 38.5 (control), 43.8 (HBOT) p = 0.54
		Week 6: 28.3 (control), 38.5 (HBOT) p = 0.25
		Month 4: 25.6 (control), 34.5 (HBOT) p = 0.33
		Year 1: 25.8 (control), 31.4 (HBOT) p = 0.53
Sarno JE, 1972,128 and Sarno MT, 1972130	100% oxygen at 2.0 atm × 1.5 hours × 1 session.	Communication baseline done 24 hours prior to exposure, and immediately after HBOT treatment. No significant effect on any measure.	Fair
New York; patients with vascular stroke at least 3 months post-stroke.	Control: 10.5% oxygen at 2.0 atm × 1.5 hours × 1 session.
32 (all received both HBOT and control; order randomized)
Rusyniak, 2003131	100% oxygen at 2.5 atm × 1 hour	No difference in proportion with good outcome on NIHSS at 24 hours (HBOT 18%, control 31%, p=0.44)	Fair
Indiana; patients presenting within 24 hours of ischemic stroke; 17 HBOT, 16 control	Control: 100% oxygen at 1.14 atm × 2 hour	90 days: No difference based on intention to treat using Barthel Index, Modified Rankin score, GOS, or NIHSS.
		Control group significantly higher proportion with “good” outcome on Modified Randkin score, GOS and NIHSS by per protocol analysis.
Nighoghossian, 1995126	100% oxygen at 1.5 atm × 40 minutes daily × 10 treatments.	Pretherapeutic and posttherapeutic differences (HBOT-Control)	Poor
France	Control: Room air at 1.5 atm × 40 minutes daily × 10 treatments.	No significant differences on Orgogozo scale, Trouillas scale, or Rankin scale at 6 months or 1 year
Patients with ischemic stroke in middle cerebral artery confirmed by CT and seen within 24 hours of onset; 17 HBOT, 17 control
Marroni, 1987129, 1988132	Group C1: 2.0 atm × 60 minutes × 30 sessions; Group C2: Same as C1 at 1.5 atm; Group D1: in-water rehab + HBOT 2.0 atm × 60 minutes; Group D2: Same as D1 at 1.5 atm; Group E1: 30 simultaneous, 60-minute HBOT + 40-minute in-water rehab sessions at 2.0 atm; Group E2: Same as E1 at 1.5 atm.	All dry HBOT groups showed greater improvement in their motor ability, but no clear-cut difference could be observed among the 4 groups that scored improvements of from 3.1 to 3.8 degrees. HBOT 1.5 atm rehab group reached 8.1 degrees 1 month after treatment, leveling off to 7.7 at 3 months, and the 2.0 atm HBOT rehab group showed an 11.6 degree improvement still present 3 months after treatment.	Poor
Italy	Control: Group A: No treatment, Group B: 30 in-water physical therapy sessions × 40 minutes.
Stabilized stroke patients no longer undergoing any form of therapy or rehabilitation who had a stroke from 2 to 172 months earlier (average 29.2 months).
80 total

The number of patients ranged from 32 to 80. Strokes were described as ischemic,^{126, 127, 131} thrombotic,¹²⁹ or vascular.^{128, 130} Two trials included only acute patients who were within 24 hours of the onset of their stroke,^{126, 131} another enrolled patients within 2 weeks of onset,¹²⁷ another enrolled only patients who were at least 2 months past their stroke (range 2 to 172 months, average 29.2 months) and were no longer receiving any therapy or rehabilitation,¹²⁹ and the last included patients at least 3 months post-stroke (range 3–108 months).^{128, 130}

Control groups were active in the randomized controlled trials; three^126–128 matched the pressure of the treatment group but used room air instead of 100 percent oxygen and one used 100 percent oxygen with 1.14 atm pressure.¹³¹ One added occupational and physical therapy to the regimen in both control and treatment group patients.¹²⁷ The non-randomized controlled trial¹²⁹ assigned 80 stroke patients to eight comparison groups with combinations of in-water or “dry” physical therapy; HBOT at different doses (1.5 or 2.0 atm); both HBOT and physical therapy; or no treatment. Patients were assigned to treatment group based on which group had an open position at the time they were assigned.

In two of the randomized trials, both patients and examiners were masked to treatment assignment.^{127, 128, 130} In the other two,^{126, 131} the title describes the study as double-blind, but there is no mention in the text about masking of outcome assessors (the patients received sham treatments). In the non-randomized controlled trial, the examiner, but not the patient, was masked to treatment assignment.¹²⁹

One study reported outcomes only immediately following a single HBOT treatment,^{128, 130} two measured outcomes at various points over 1 year,^{126, 127} and two followed patients for 3 months.^{129, 131}

Mortality. Only one pilot study of HBOT for stroke within 24 hours of onset of symptoms reported mortality, finding two deaths in the sham group (12.5 percent) and one in the HBOT group (6 percent) at 3 months.¹³¹ However, the study sample size was too small to detect a difference in mortality, and the causes of death were not clearly reported.

Neurological outcomes. Anderson, Bottini, et al. (1991), a fair-quality, double-masked, randomized controlled trial that enrolled patients within 2 weeks of the onset of their stroke, found no significant differences between control and HBOT groups on graded neurological exams at day 5, week 6, month 4, or year 1 of followup.¹²⁷ Patients received up to 15 HBOT sessions every 8 hours at 1.5 atm. Treatment was not well accepted by patients. Twenty percent dropped out before completing the study, and 38 percent deviated from the protocol in some way. Although the differences were not statistically significant, the study was suspended early because the improvements were consistently greater in the control group.

This was the only controlled trial that reported the number of patients who were screened for inclusion. Less than half of those screened were enrolled (39/92 patients screened). Patients were excluded if they had medical contraindications for HBOT, were over age 90, had a score of less than 20 on a graded neurological exam, had deficits that were rapidly improving, or had a treatment status of “supportive care only.”

Sarno, Rusk, et al. (1972), another fair-quality, crossover trial, found no difference in communication and cognitive outcomes in 32 stroke patients who received a single HBOT session or sham treatment.^{128, 130} The main strength of this trial is that patients and outcome assessors were masked to treatment allocation, and the study used a battery of standardized tests, so the results are less likely to be affected by observer bias. The conclusions that can be drawn from this study are limited, however, because it involved only a single session of HBOT. The investigators reported that they were able to enroll only 53 percent of the original number of patients planned because of difficulty recruiting subjects. They stated that some patients refused to participate when they heard of a lack of effect of treatment. It is unclear how potential participants would have received this information before the conclusion of the study, and it raises the possibility that patients who participated may have been different from the general population of eligible subjects.

The third randomized controlled trial, Nighoghossian, Trouillas, et al. (1995), ¹²⁶ enrolled 34 patients within 24 hours of the onset of their stroke. Twenty-seven (79 percent) patients completed the study. When the mean neurological scores of the groups were compared, patients who received 10 HBOT treatments had significantly higher scores at 12 months on two of the three scales used (Orgogozo scale and Trouillas scale) compared to patients who were assigned to the control group. Mean scores did not differ on these scales at 6 months or on the third scale used (Rankin). Baseline Orgogozo scale scores in the HBOT group were higher, and when these differences in baseline measures were taken into account in the data analysis by comparing the mean change from baseline to 6 months, no significant differences were found on this scale. Baseline scores on the Rankin and Trouillas scales are not reported, but there was no difference in the mean change from 6 months to 12 months on either.

This study was rated poor-quality because there was a difference in the neurological scores of the groups. Because randomization and allocation concealment methods were not described, we cannot assess whether randomization was appropriately conducted, but the significant difference in baseline prognostic factors suggest that randomization failed to result in comparable groups.

Rusyniak et al. (2003) was a fair quality, randomized pilot study that enrolled 33 patients within 24 hours of stroke to a single session of HBOT or a sham treatment. This study assessed the proportion of patients with “good” outcome at 24 hours, based on the National Institutes of Health Stroke Scale (NIHSS), and 90 days on the NIHSS, the Barthel Index, the Modified Rankin score, and the Glasgow Outcome Scale. Good outcome was defined as a score of zero or and improvement of greater than 4 points on the NIHSS from baseline, a score of 95 or 100 on the Barthel Index, a score </= 1 on the Modified Rankin score, and a Glasgow Outcome Scale score of 5. Three patients were lost to follow up in the control group, but an intent-to-treat analysis is presented. At 24 hours 32 percent of control patients had a good outcome compared to 18 percent in the HBOT group (p = 0.44). At 90 days, there were no significant differences seen on any measure based on intention to treat analysis, although the proportions of patients with good outcome were higher in the control group for all measures. Using a per-protocol analysis, including only patients who completed 90 days of follow-up, the control group had significantly more patients with good outcome based on the NIHSS, Modified Rankin Scale and the Glasgow Outcome Scale.

The last controlled study, Marroni et al. (1987), was a non-randomized trial that used eight different treatment regimens combining HBOT with in-water or dry physical therapy compared to no treatment or in-water physical therapy alone (no HBOT).^{129, 132} The number of patients in each treatment group ranged from 7 to 12; all were stable and were no longer receiving any therapy or rehabilitation. Mean outcome measure scores were plotted for each group. After 60 days, the groups of patients treated with HBOT improved by 1 and 1.8 degrees on the Kurtzke functional scale (a scale measuring walking ability and other abilities in patients with multiple sclerosis), while control groups had no improvement. Groups receiving HBOT plus physical therapy improved more, and the in-water HBOT group achieved the largest improvement. Patients were also evaluated on the Neuromotor Disabilities Evaluation Scale. This unvalidated scale, developed by the study authors, measures 10 groups of limb and system function (e.g., finger and hand function, muscular strength, walking ability) on a scale ranging from 17 (best) to 111 (worst). Over a 3-month evaluation period, mean scores in the dry HBOT groups improved by 3.1 to 3.8 degrees, and patients in control groups improved 1 degree. There were no differences among these groups based on concurrent physical therapy or HBOT dose (1.5 or 2.0 atm). The groups receiving HBOT and concurrent in-water physical therapy improved by 7.7 degrees (1.5 atm) and 11.6 degrees (2.0 atm).

This study, which we rated poor-quality, provided no information by which to judge the comparability of the groups at baseline. For this reason it is impossible to rule out bias or confounding as explanations for the results. Because it combined HBOT with in-water physical therapy, it is impossible to determine which component was responsible for the reported improvements. Because patients were not allocated to treatments randomly, the investigators could knowingly or unknowingly assign patients with a better prognosis to a treatment they believed in. Likewise, the patients and outcome assessors were not masked to treatment; the patient could have altered their level of participation, and the outcome assessor could have knowingly or unknowingly interpreted outcomes differently, depending their beliefs about the treatment received.

Observational Studies

Two studies reported mortality rates; ^{140, 147} three measured grip strength with a dynamometer; ^{134, 143, 144} one performed a mental status examination, two-point discrimination, and repetitive thumb/finger movements; ¹⁴⁴ two measured spasticity on a five-point scale; ^{134, 143} and one measured 33 different functions of cognition and motor ability. ¹³⁷ One study¹⁰⁸ used a scoring system that included one standard test to measure memory (Bender-Gestalt Memory Test), but the other components of the score were not validated or well described.

In three studies,^{108, 143, 144} outcomes were measured at the conclusion of HBOT treatment. Since the duration of treatment varied according to patient response, the timing of these followup measures also varied. Others followed up patients for 6 weeks, ¹⁴¹ every 3 months for 1 year, ¹⁴⁸ at 6 and 12 months after treatment, ¹⁴¹ and 4.5 years after treatment.¹⁴⁶

All of these studies had fatal flaws that led to a poor-quality rating. Only two studies established that all patients were stable at the time baseline measures were taken.^{137, 139} In a before-after study, we can be confident that the results are due to treatment and not to the natural course of the illness only if the stability of their baseline condition is established clearly. This was not the case in any of these studies.

In two studies, some of the patients were observed for a long enough period to establish that they were stable.^{137, 139} However, both these studies had other flaws that led to a poor-quality rating. Both used simultaneous co-interventions, such as physical therapy and biofeedback, making it impossible to determine the effects of HBOT alone. Also, they used subjective outcomes and did not mask the outcome assessors to knowledge that patients had received HBOT treatment. Masked outcome assessment is especially important when outcomes are subjective or depend on the judgment of the assessor.

As a group, the observational studies reported that between 20 and 83 percent of selected patients with stroke improved after HBOT therapy.

Mortality. A retrospective comparison of cohorts examined 5-year mortality rates in 65 patients who received HBOT compared to 65 patients who did not receive HBOT.¹⁴⁷ Thirty-two percent of the patients who received HBOT died, compared to 48 percent of those who did not receive HBOT (p < 0.05). There is no information on how patients were chosen for inclusion, and although patients were matched by age, sex, and time of event, significant differences in clinical history between the two groups existed (HBOT-treated patients had more hypertension, respiratory insufficiency, and vascular insufficiency of the inferior limbs). The groups were treated at two different hospitals, and it cannot be ruled out that factors other than HBOT treatment accounted for the reduction in mortality in the HBOT group. For example, 10.8 percent of patients in the HBOT-treated group received diuretics, compared with 4.6 percent of the control group patients (p < 0.05).

In a time series, 40 patients¹⁴⁹ who were at least 4 weeks from the onset of their ischemic attack and whose neurologic state was unchanged for at least 3 weeks were given HBOT treatment for 15 days, and then after a rest of 30 days, 15 more treatments if they had improved. Seven patients (17.5 percent) died. The length of the followup is unclear, so we cannot compare these results to other reports. This study was poor-quality, because stability at baseline was not established, outcomes were subjective, outcome assessors were not masked, and other interventions were given.

Neurological examinations. Uncontrolled observational studies found that the majority of patients showed at least some improvement, and some had extremely good results. The proportion of patients whose improvement on unspecified neurological measures was “marked,” “excellent,” “dramatic,” or “completely cured” ranged from 6 to 64 percent. Improvement was “good” or “moderate” in 20 to 82 percent of patients. In one time series, 82.5 percent of 40 patients improved, and 11 of 15 (73 percent) of patients with aphasia improved.¹⁴⁹ The outcome measure used was not described. Two studies^{133, 145} reported initially good responses (48 to 67 percent) during or immediately after HBOT treatment, but the improvement was maintained at followup in only 8 or 9 percent of patients. It is difficult to draw conclusions from these studies because the outcome measures are described in vague terms, and baseline measures were not presented.

Other outcome measures. Three studies measured grip strength using a dynamometer,^{134, 143, 144} but they did not report results of all patients, only example cases. Two studies by the same author measured spasticity on a five-point scale.^{134, 143} All patients also received physical therapy in addition to HBOT treatment. All patients with spasticity improved rapidly during HBOT treatment. Although the improvement was transitory, it was prolonged if physical therapy was performed in the chamber during HBOT. Improvement was maintained in “many” patients (number not specified) at 3 to 12 months of followup.

A before-after study of 50 patients measured 16 self-reported functions and 33 functions reported by physical therapists before and after HBOT therapy.¹³⁷ Following treatment, 96.7 percent of patients reported total improvement on at least one function, and 3.3 percent reported no improvement. Physical therapists reported 82 percent of patients showed good or excellent improvement on at least one function. Sixty-seven percent of patients rated the program “excellent” or “stupendous.” Outcome assessors were not masked, and outcome measures were subjective and not defined.

A before-after study of 122 patients¹³⁹ reported that the degree of improvement did not appear to be related to the patient's condition at baseline, whether initiated when the patient was bedridden (55 percent improved), wheelchair bound (71 percent improved), or walking with aids (56 percent improved). Although some patients in this study were as many as 10 years post-stroke, the possibility that these results were due to bias, confounding, or both cannot be ruled out. Because the outcome measures used were subjective and the outcome assessor was not masked, the results may be biased. It is also not possible to determine if the results are due to the simultaneous physical therapy that was used in some patients.

Abstracts and Conference Proceedings

One of the controlled trials used a Neurological Recovery Score (improvement at 12 months HBOT vs. control, p = 0.031); the other reported recurrent stroke or TIA (4.8 percent TIA in HBOT vs. 5.9 percent stroke in control, p not given).

Two uncontrolled studies reported observations or results of unspecified neurological examinations. One stated that 100 percent of 18 patients with chronic traumatic, ischemic, hypoxic, and anoxic brain injuries showed motor, behavioral, personality, or cognitive gains by 40 treatments. In another, 80 percent of 140 patients with ischemic stroke improved. The other did not report the proportion of patients who improved, but reported that “nearly all” patients who responded favorably to HBOT showed a positive response to extra-intracranial arterial bypass surgery. Other uncontrolled studies measured hand grip and spasticity¹⁴³ (improvement in all four patients), recovery of consciousness (17 percent of six patients regained consciousness), and short-term memory quotient (memory quotient improved significantly from baseline, p < 0.001).

Synthesis

The best available evidence shows no benefit from HBOT for stroke, but there are conflicting results from flawed controlled trials and uncontrolled studies, and no good-quality study has been conducted. Fair-quality randomized trials found no benefit in patients treated with HBOT over patients treated with pressurized room air or low-pressure oxygen. Two of these trials have limited applicability because they evaluated only a single HBOT treatment. Two flawed trials found that HBOT improved neurological outcomes on some measures. There was no pattern among these studies to suggest that any particular dose or frequency of HBOT treatment is more effective than any other.

Most observational studies reported good, and sometimes dramatic, results, but failed to prove that these results can be attributed to HBOT. Design flaws make it impossible to rule out other explanations for their results. Failure to establish that patients were stable at baseline, and the use of other treatments, made it impossible to separate out the effects of HBOT alone. Finally, the lack of masking of outcome assessors and the use of subjective outcome measures in most studies make it impossible to rule out bias in outcome assessment. Therefore, no conclusions about the effectiveness of HBOT for stroke can be drawn from this body of evidence.

Neurological Complications

Pulmonary Complications

HBOT can cause aspiration, which may cause pneumonia or increased oxygen requirements. Patients who have a reduced level of consciousness and those who have gastroesophageal reflux have a higher risk of aspiration. Aspiration results when swallowed air leads to distension of the stomach and, consequently, regurgitation of stomach contents into the pharynx, where they may be drawn into the trachea and lungs.

Paralysis, lack of control, or weakness of the muscles involved in the gag reflex or swallowing are also risk factors for aspiration. Placing a feeding tube into the stomach can reduce the risk of aspiration. This prevents distension of the stomach and reduces the risk of regurgitation.

Pulmonary complications were relatively common in the trials of brain-injured patients. In the Artru trial, which used 2.5 atm, treatment was stopped in 35 percent (11/31) of sessions due to pulmonary symptoms.⁸⁸ No further information was provided about the severity of these complications, their duration, or whether treatment was restarted later. The chamber type was not reported. In the Rockswold trial, which used 1.5 atm, HBOT had to be permanently stopped because of adverse effects in 10/84 (12 percent) patients. The reasons for withdrawal of therapy were not described clearly. Pulmonary complications (increasing FiO2 requirement and/or infiltrates detected on chest x-ray) were described as the most frequent complication, but the number of cases was not reported.

There are no reliable data on the incidence of aspiration in children treated for cerebral palsy with hyperbaric oxygen. No cases of aspiration were reported in the Collet and Cornell trials.^118–120 Nuthall reported two cases of children with cerebral palsy who, after HBOT, had acute respiratory failure requiring ICU admission.¹⁵² The first child underwent two HBOT dives at 1.75 atm. Between the two dives he was fed a meal, which he aspirated during the second dive. The authors noted that, during the hyperbaric treatments, the child's head was completely enclosed by a vinyl hood with latex seals around his neck for the delivery of oxygen. The second child was thought to have suffered an air embolism.

In response, Harch, Deckoff-Jones, and Neubauer wrote a letter to the editor of the journal Pediatrics. They noted that the incidence of pulmonary aspiration in practice is unknown, but, between them, they had “logged over 35,000 treatments on brain injured children without a single case of primary aspiration or air embolism.”¹⁵³

Ear Problems

HBOT can cause pain, rupture, or hemorrhage in the ear. The most common symptom is a feeling of pressure and pain in the eardrum. Alert patients can prevent these symptoms by swallowing and other maneuvers (similar to what one does when the pressure in the cabin of a plane increases during a descent.) In children and other patients who cannot perform these maneuvers, myringotomy (making a hole in the eardrum) can prevent the symptoms.

In acute TBI, the Rockswold study, which used 1.5 atm, reported that after two of 38 patients (5 percent) had hemotympanum, prophylactic myringotomies were performed on the last 46 patients enrolled in the HBOT arm. ^89–91 A monoplace chamber was used in this study.

The controlled trial of HBOT for cerebral palsy (Collet et al) provided the best data on the frequency of ear problems in children. In that trial, 47 percent of children assigned to HBOT and 22 percent of children assigned to compressed air developed ear problems due to compression. Both multiplace and monoplace chambers were used in this study, but the data were not stratified by this variable. In the Cornell study (using 1.5 atm pressure), 35 percent of patients experienced ear problems related to pressure.¹¹⁸ Chamber type was not reported in this study. Ear problems were not reported in the controlled studies of stroke patients, and one uncontrolled study reported ear problems in 6 percent of 122 patients, with 1 percent requiring myringotomy.¹³⁹

Quality of Evidence about Adverse Events

Except as noted above, the quality of evidence about adverse events was almost universally poor in the studies we reviewed. Some studies reported overall rates of withdrawal, but did not specify whether patients withdrew from the HBOT or the control group. Most did not state the reasons for withdrawal. None of the observational studies described a protocol for detecting and classifying adverse effects. Without such a protocol, event rates are almost certain to be underestimated.

Traumatic Brain Injury

Overall, the two available fair-quality trials provide fair evidence that HBOT might reduce mortality or the duration of coma in severely injured TBI patients. However, in one of these trials, HBOT also increased the chance of a poor functional outcome. Therefore, they provide conflicting evidence to determine whether the benefits of HBOT outweigh the potential harms.

Although they are cited frequently, the case series and time-series studies of HBOT for TBI patients had serious flaws. There were no high-quality studies of the use of HBOT to improve function and quality of life in patients with chronic, stable disabilities from TBI. The most important gap in the evidence is a lack of a good quality time-series study or controlled trial of the effects of HBOT on cognition, memory, and functional status in patients with deficits due to mild and moderate chronic TBI.

Studies of the effects of HBOT on ICP levels also had mixed results. HBOT may be effective in reducing elevated ICP in some acute TBI patients, but rebound elevations higher than pretreatment levels can occur. The clinical benefit of the ICP lowering and the harm attributable to the rebound elevations are unclear. Without further delineation of the patient or treatment factors that may be associated with successful lowering of ICP, the current evidence is insufficient to determine whether the overall effect of HBOT on ICP is beneficial or harmful.

Other Brain Injury

We did not identify any good or fair-quality studies of HBOT for anoxic-ischemic encephalopathy. We found one randomized controlled trial and five before-after studies of patients with various kinds of nontraumatic brain injuries. All of these studies were poor-quality. The controlled trial lacked details regarding the subjects' recruitment and baseline characteristics and the methods used to randomize subjects and measure outcomes. All five before-after studies lacked objective outcome measures and masked assessment, and timing of baseline and followup measures was not clear.

Randomized Controlled Trials and Cohort Studies

Seven categories of criteria apply to RCTs and cohort studies. They include:

1. Initial assembly of comparable groups.

   1. For RCTs: adequate randomization, including first concealment and whether potential confounders were distributed equally among groups.

   2. For cohort studies: consideration of potential confounders with either restriction or measurement for adjustment in the analysis; consideration of inception cohorts.

2. Maintenance of comparable groups (includes attrition, cross-overs, adherence, contamination).

3. Levels of follow-up: differential loss between groups; overall loss to follow-up.

4. Measurements: equal, reliable, and valid, and including masking of outcome assessment.

5. Clear definition of interventions.

6. All important outcomes considered.

7. Analysis:

   1. For RCTs: intention-to-treat analysis.

   2. For cohort studies: adjustment for potential confounders.

The definitions of the three rating categories for these types of studies are as follows:

Good: Comparable groups assembled initially and maintained throughout the study; follow-up at least 80 percent; reliable and valid measurement instruments applied equally to the groups; outcome assessment masked; interventions defined clearly; all important outcomes considered; appropriate attention to confounders in analysis; for RCTs, intention-to-treat analysis.

Fair: Generally comparable groups assembled initially but some question remains whether some (although not major) differences occurred with follow-up; measurement instruments acceptable (although not the best) and generally applied equally; outcome assessment masked; some, but not all important, outcomes considered; appropriate attention to some, but not all, potential confounders; for RCTs, intention-to-treat analysis.

Poor: Groups assembled initially not close to being comparable or not maintained throughout the study; measurement instruments unreliable or invalid or not applied at all equally among groups; outcome assessment not masked; key confounders given little or no attention; for RCTs, no intention-to-treat analysis.

Aggregate Internal Validity

This category refers to the overall extent to which data are valid for conditions addressed within studies. It would be rated according to quality grading information about individual studies.

Aggregate External Validity

This category concerns the generalizability of evidence to questions addressed by the linkage. This would include the concordance between populations, interventions, and outcomes in the studies reviewed (on the one hand) and those to which the linkage pertains (on the other). In short, this category reflects the applicability of the evidence to real-world conditions.

The Methods Work Group expects that differences between conditions examined in studies and those addressed by the linkages should be considered if they could potentially influence outcomes. These might include (but not necessarily be limited to): (a) biologic or pathologic characteristics; (b) incidence and prevalence of clinical conditions; (c) distribution of comorbid conditions that might affect outcomes; and (d) likelihood of acceptability and adherence on the part of patients or providers (or both).

Consistency

This category relates to the overall “coherence” of the body of evidence relating to the linkage. Specifically, it includes the number of studies, the homogeneity of those studies (in terms of clinical conditions, populations, settings, and the like), the level of precision of findings in the studies, and the direction of results. In addition, it can include dose-response relationships.

Experimental Studies

1. Was the assignment to the treatment groups really random?

   • Adequate approaches to sequence generation

     • Computer-generated random numbers

     • Random numbers tables

   • Inadequate approaches to sequence generation

     • Use of alternation, case record numbers, birth dates or week days

2. Was the treatment allocation concealed?

   • Adequate approaches to concealment of randomization

     • Centralized or pharmacy-controlled randomization

     • Serially-numbered identical containers

     • On-site computer based system with a randomization sequence that is not readable until allocation

     • Other approaches with robust methods to prevent foreknowledge of the allocation sequence to clinicians and patients

   • Inadequate approaches to concealment of randomization

     • Use of alternation, case record numbers, birth dates, or week days

     • Open random numbers lists

     • Serially numbered envelopes (even sealed opaque envelopes can be subject to manipulation)

3. Were the groups similar at baseline in terms of prognostic factors?

4. Were the eligibility criteria specified?

5. Were the outcome assessors blinded to the treatment allocation?

6. Was the care provider blinded?

7. Was the patient blinded?

8. Were the point estimates and measure of variability presented for the primary outcome measure?

9. Did the analyses include and intention-to-treat analysis?

Observational Studies

Cohort studies:

• Is there sufficient description of the groups and the distribution of prognostic factors?

• Are the groups assembled at a similar point in their disease progression?

• Is the intervention/treatment reliably ascertained?

• Were the groups comparable on all important confounding factors?

• Was there adequate adjustment for the effects of these confounding variables?

• Was a dose-response relationship between intervention and outcome demonstrated?

• Was outcome assessment blind to exposure status?

• Was followup long enough for the outcomes to occur?

• What proportion of the cohort was followed up?

• Were dropout rates and the reasons for dropout similar across intervention and unexposed groups?

Case series:

• Is the study based on a representative sample selected from a relevant population?

• Are the criteria for inclusion explicit?

• Did all individuals enter the survey at a similar point in their disease progression?

• Was followup long enough for important events to occur?

• Were outcomes assessed using objective criteria or was blinding used?

• If comparisons of sub series are being made, was there sufficient description of the series and the distribution of prognostic factors?

Values of Patients and Caregivers

Values of Clinicians

Patterns and Themes

There were four overarching themes that emerged from the data: 1) View of Medicine, 2) Attitudes toward Risk, 3) Hope, and 4) Politics.