Controversy surrounds the role of steroids in the treatment of intracerebral haemorrhage.1 Theoretically, the short term use of dexamethasone is justified because it lessens the damaging effects of cerebral oedema, decreases intracranial pressure, and strengthens the blood-brain barrier. However, the possible benefit has to be weighed against the risk of complications, especially infections and gastrointestinal haemorrhage.2

So far, only two randomised, controlled trials concerning this controversy have been reported. One trial of 40 patients assumed to have intracerebral haemorrhage, found no beneficial effects of steroids.3 However, this study had 22 patients with haemorrhagic infarcts or posterior fossa haemorrhage, and the outcome measures used had little clinical relevance regarding functional ability of the patients. The other trial, from Thailand,4was well designed, but had to be terminated prematurely after an interim analysis disclosed lack of benefit and presence of clinically important adverse effects. Patients in the dexamethasone group had more frequent infections, gastrointestinal haemorrhage, and diabetogenic effect, than the placebo group. There was a possible longer early survival in a subgroup with less severe stroke. Good recovery was noted in 17% of patients in the dexamethasone group, compared with 10% of patients in the placebo group, giving a difference of 7% in favour of the dexamethasone group.

In our experience, the complication rates are not as high with dexamethasone as reported by Poungvarin et al.4 To consider the issue of safety of dexamethasone in primary supratentorial intracerebral haemorrhage, we undertook a double blind, randomised, placebo controlled trial as a pilot project.

Twenty six patients in the age group 40–80 years, with primary supratentorial intracerebral haemorrhage confirmed by brain CT, presenting within 5 days of onset were included. Patients with a history of previous disabling stroke or contraindications to steroid treatment were excluded. Informed consent was taken from relatives of all patients before admission into the trial. Dexamethasone was given intravenously, at a dose of 4 mg 6 hourly for 12 days, followed by 4 mg 12 hourly for two days and 2 mg 12 hourly for 2 days. Placebo injections of saline were given in the same dosage, from ampoules of similar size and shape, and with comparable labels, filled with a colourless solution indistinguishable from dexamethasone. No one involved with the study knew whether a particular patient was receiving dexamethasone or placebo. All patients received injectable ranitidine, 50 mg 8 hourly, for the period of the trial. The remainder of patient care was specified by their attending neurologist. The allocated treatment was stopped if the attending neurologist thought that the patients had developed a complication likely to be caused by, or aggravated with dexamethasone. No other antioedema measure (mannitol/glycerol) or neurosurgical intervention was undertaken, except in one patient who was found to have a ruptured anterior communicating artery aneurysm, diagnosed after randomisation.

Details of history and physical examination conducted at the time of admission were recorded. Haematoma size was calculated using the formula given by Kothari et al.5 Outcome measures of all patients were assessed by one of us (PD), using the Glasgow outcome scores6 at day 7 and at discharge. Fasting blood sugar was done one week after starting the treatment, to detect any diabetogenic effect. The development of fever after entry into the study was considered to be indicative of infection, irrespective of whether the focus of infection was detected or not. Statistical methods used were χ² test with continuity correction for categorical variables and a two tailed t test for continuous data.

The clinical characteristics, complications, and outcome of the patients are shown in the table. Both groups were well matched for age, hours from haemorrhage to treatment, blood pressure, volume and location of haematoma, and presence of intraventricular extension of blood. However, there were more comatose patients (Glasgow coma scale score<7) in the dexamethasone group (5/12) than in the placebo group (3/14).

The complication rates were higher in the placebo group than in the dexamethasone group (placebo group, 10 v dexamethasone group, 4), but the difference was non-significant. One patient in each group was found to be non-eligible after randomisation, and their allocated treatment was stopped within 3 days, but they continued to remain in the allocated group for analysis.

In all, seven patients died (five patients in the dexamethasone group died due to herniation within the first week and two patients in the placebo group died; one died of herniation in the first week, and the other died due to septicaemia on day 36). There was no difference between the groups in the number of patients who had a good outcome (2/12 in the dexamethasone group v 2/14 in the placebo group).

There could be many reasons why Poungvarin et al 4 found a higher frequency of adverse effects due to dexamethasone. Firstly, it may have been a chance finding. If they had continued the study further, the placebo group may have had similar complications such as infections or hyperglycaemia, thereby making the difference between the two groups non-significant. Secondly, the dexamethasone group may have had a higher proportion of serious patients who are more likely to develop complications such as infections, gastrointestinal haemorrhage, or hyperglycaemia. This may occur despite stratified randomisation, particularly when the numbers are few in the individual stratum. Arguably, similar factors in our placebo group may explain our findings. This may be true, but it emphasises the few and hence not very reliable data on whether dexamethasone is too unsafe to be used.

We conclude that dexamethasone in the regimen used in our trial is not likely to produce an unacceptably high rate of adverse effects in patients with primary supratentorial intracerebral haemorrhage. Whether or not dexamethasone or other corticosteroids are beneficial in this setting needs further study. The study by Poungvarin et al,4 points to a higher proportion of patients making complete recovery by the 21st day in the treatment group (odds ratio 0.7, 95% CI, 0.2–2.4), this needs to be either confirmed or refuted.