EDITOR—We would like to clarify some of the statements that Weaver made about our double blind randomised trial of hyperbaric compared with normobaric oxygen treatment of carbon monoxide poisoning. 1 2 Weaver expresses concern that concomitant depression and use of psychoactive drugs might have influenced the results, given the large percentage of suicide attempts in our cohort of patients. Depression and the use of drugs may indeed have resulted in a higher incidence of poor outcome overall, but this would not have biased the comparison between normobaric and hyperbaric groups: patients were specifically stratified for attempted suicide before randomisation to treatment.

Weaver seems to be concerned about the delay in receiving hyperbaric oxygen treatment. Although the geometric mean treatment delay was 7.1 hours, we performed subgroup analysis of patients treated within four hours (all patients, and severely poisoned patients alone). We also analysed outcome in four groups according to time to treatment (<3, 3-60, 6-12, >12 hours) and found no difference in outcome between hyperbaric and normobaric oxygen groups. Further multivariable analysis did not identify delay in treatment as a predictor of poor outcome. Thus there was no evidence that delay to treatment might have explained the lack of benefit of hyperbaric oxygen.

Weaver also questions our use of cluster randomisation. With this we allocated more than one person simultaneously to the same treatment on 22 occasions (two on 12 occasions, three on five occasions, and four on five occasions). Overall, 14 clusters (40 patients) were allocated to hyperbaric oxygen and eight clusters (19 patients) to normobaric oxygen. Continuous outcome variables were analysed by the mixed procedures in SAS, which allows a repeated measures analysis of variance, with the variable cluster being treated as a random repeated measurement, thus adjusting for within cluster variation.3

We also repeated the analysis excluding all patients who were allocated as part of a cluster and further repeated it adjusting for the three variables (exposure time, time to carboxyhaemoglobin measurement, and time to treatment) that uniquely defined the cluster. These analyses suggest that our results were not biased by cluster randomisation.

Weaver refers to Thom et al's findings of no relapses in their patients given hyperbaric oxygen, but it is important to note that all five relapses in our study occurred in patients given hyperbaric oxygen (P=0.03).

EDITOR—Weaver presents a well balanced editorial on the controversy surrounding the treatment of carbon monoxide poisoning with hyperbaric oxygen.1 This is refreshing, as a previous editorial in the BMJ was not so objective.2

Weaver mentions four prospective randomised studies of normobaric versus hyperbaric oxygen. The latest of these is the most convincing, being prospective, randomised, and double blind, with sham hyperbaric treatments.3 Weaver is critical of the use of continuous high oxygen concentrations for three days in the control group because this was not representative of usual practice. Previous studies have been flawed by a failure to optimise treatment in the normobaric group, and Scheinkestel et al's study clearly shows that such cheap, available, safe treatment is also effective. The claim that optimal normobaric oxygen treatment is not in routine use for this condition is cause for considerable concern.

Weaver also criticises Scheinkestel et al for not applying hyperbaric oxygen earlier in their study. This criticism comes despite the lack of evidence from controlled prospective comparative studies that earlier treatment is any more beneficial and despite the fact that subgroup analysis of treatment within four hours showed no benefit from hyperbaric oxygen. Scheinkestel et al's study is representative of most clinical practice because of late presentation and the need for stabilisation and transport to a remote hyperbaric facility.

Scheinkestel et al's study has shown that hyperbaric oxygen results in a worse outcome than does normobaric treatment. Even if it is wrong, the degree of any benefit is unlikely to be clinically important compared with the risk of such treatment. Hyperbaric oxygen and the associated transportation are associated with appreciable hazards to both the attendant and the patient, which are often understated. I therefore suggest that the multicentre study that Weaver proposes would now be unethical.

In carbon monoxide poisoning 100% oxygen should be given immediately and continued for several days. Resources should be concentrated on promulgating this message together with preventing carbon monoxide poisoning and detecting it early rather than on providing more hyperbaric oxygen facilities. The NHS and health- care agencies in the United States should review their funding strategy for the use of hyperbaric oxygen in acute carbon monoxide poisoning, because the only benefit would seem to be towards the profitability of independent hyperbaric facilities.

EDITOR—Scheinkestel et al clarify the issues regarding concomitant depression, delay to hyperbaric oxygen, and cluster randomisation in their clinical trial.1 I agree that attempted suicide probably did not bias the outcome between the two arms.2 The data provided strengthen the inferences from their trial.

Thom et al found no delayed neuropsychological sequelae in their patients given hyperbaric oxygen,3 whereas Scheinkestel et al found relapses only in patients given this treatment. Clearly the evidence regarding the effectiveness of hyperbaric oxygen in carbon monoxide poisoning remains conflicting.

In the United States the commonest treatment for acute carbon monoxide poisoning is inhalation of oxygen by high flow, non-rebreathing face mask (70-80% fractional inspired oxygen), or 100% oxygen if the patient needs intubation, for 4–6 hours. I am unaware of anyone who treats acute poisoning with 100% oxygen for 2–3 days, apart from Scheinkestel et al in their trial. No trial has shown that inhalation of normobaric oxygen improves outcome in carbon monoxide poisoning or the optimal duration of this treatment.

Since the neuropsychological and disability rate in Scheinkestel et al's control group was relatively high and similar to that seen by others,3^–7 it is unclear if giving 100% normobaric oxygen for three days has any advantage over giving 70-100% oxygen for only a few hours. Since optimal oxygen treatment is poorly defined, I disagree with Tighe that three days of treatment is indicated for acute carbon monoxide poisoning.

Tighe takes exception to the delay to hyperbaric oxygen treatment, and Scheinkestel et al have provided illustrative additional information. I agree with Tighe that transportation to a hyperbaric oxygen chamber can present problems. I disagree with him that a multicentre prospective randomised controlled trial of hyperbaric oxygen would be unethical. On the basis of existing trials, 1 3 4 7 such a trial would be ethical because present evidence is conflicting. Because the evidence is conflicting I see no reason why hyperbaric oxygen treatment for acute carbon monoxide poisoning should be abandoned.

The NHS and the Health Care Financing Administration in the United States might review their funding strategies for hyperbaric oxygen treatment in acute carbon monoxide poisoning, but the cost of hospital admission for three days (to provide 100% normobaric oxygen) is also considerable and needs to be considered. Regardless of the treatment of carbon monoxide poisoning, increased societal awareness and prevention of carbon monoxide poisoning remain critically important.