Since the 1970s, the focus of mental health policy in the UK has been on those with severe, enduring psychotic illness, relocating the care for such people from the hospital to the community. This has now been largely achieved. The policy focus is now shifting to the other end of the process, establishing specific services for those with early psychosis, as flagged in the NHS National Plan. Spencer et al describe the principles underlying the Birmingham Early Intervention Service. This is the first and best developed such service in the UK, its ethos based on the work evaluating early intervention and relapse prevention done by Max Birchwood and colleagues.

The principles underlying a comprehensive first-episode service are cited by Spencer et al: early detection; early, effective working to normalise individuals' experience of psychosis; and relapse prevention.

Early detection aims to reduce the duration of untreated psychosis (DUP), a well-replicated and surprisingly strong predictor of short- and long-term outcome. Average or median DUP in most systematic UK studies tends to be shorter than Spencer et al suggest, at least for positive symptoms. However, this does not mean that working to reduce it is any less important. Drake et al (2000) found the median duration to be 12 weeks in a large sample of first-admission patients. The longer the DUP, the more severe was the symptom level at baseline, but even controlling for this effect, DUP was the single strongest predictor of symptomatic outcome at 3 months. Drake et al's study modelled DUP against short-term outcome and found a non-linear relationship. This suggests that bigger gains in improving outcomes would be made by bringing forward treatment by 1–2 weeks in cases of relatively short DUP than by bringing forward treatment by 1–2 months in those with long DUP. In this respect, adjustments to service configuration, such as assessing all new cases of psychosis within 1 day, should have an impact. Further steps include training general practitioners; targeting high-risk groups such as those with a family history; and even high-profile public education campaigns for whole communities. Using all these approaches together (details available from S.L. upon request) showed a reduction in median DUP in a Norwegian county from 26 to 5 weeks, with a corresponding reduction in how severely ill patients were when they presented.

The next step is early, effective and acceptable treatment. Antipsychotic drug treatment remains a central part of the management of the first episode. Which class of drug to choose and which individual drug within that class is still a matter of debate. What is clear is that people in their first episode are sensitive both to the therapeutic effects and the side-effects of drug treatment. As a class the new atypicals offer a lower incidence of neurological side-effects than do the conventionals, although speed of remission in the first episode has been shown to be the same for conventionals and atypicals. Advantages of new atypicals are more likely to emerge in the longer-term prevention of relapse.

Offering a conventional drug first line is still a legitimate strategy, with two caveats. First, there is increasing evidence that low dosages are effective in treatment-na¨ve patients. From the viewpoint of basic pharmacology, doses equivalent to as little as 2 mg haloperidol daily will give 80% dopamine D2 receptor blockade. In a trial in which clinicians titrated haloperidol dose blindly against clinical response in 130 first-episode patients, the mean dose at the end of 12 weeks' treatment was found to be about 4 mg daily, with only 4 % of patients needing more than 6 mg (details available from S.L. upon request). Given its high rate of extrapyramidal side-effects (EPS), it may be that haloperidol is not the best conventional to choose in the first episode. Its high potency means it has a narrow therapeutic range between the dose needed to work and the dose at which EPS will emerge.

The second caveat if using a conventional drug first line is to assess and monitor symptoms and side-effects regularly and to change to an atypical the moment poor response or side-effects emerge. Rating scales to assess these issues are underused. Health commissioners will be much happier to fund drugs they view as expensive if clinicians can show they are monitoring outcomes and side-effects objectively.

Data are emerging that point to the effectiveness of adjunctive cognitive-behavioural therapy (CBT) in accelerating remission from acute symptoms in the first episode. In a large trial, 316 patients, of whom 80% were first admissions, were randomised a mean of 6 days after admission to receive a 5-week package of CBT in addition to routine care, or a 5-week package of supportive counselling in addition to routine care, or routine care alone. Assessments blind to treatment allocation showed that the group receiving CBT improved significantly more rapidly over 50 days in terms of delusions and hallucinations than did the other groups (details available from S.L. upon request).

The real challenge in longer-term management is to prevent relapse. Data from Lieberman's group (Reference Robinson, Woerner and AlvirRobinson et al, 1999) and others remind us that, whereas over 80% of cases of first-episode schizophrenia will achieve good remission at an average of 3 months after starting treatment, one in six of these will have relapsed within 1 year and, by 5 years, five out of six will have relapsed. Moreover, cohort studies have confirmed how the speed and quality of remission are progressively impaired after the first and subsequent relapses. Relapse prevention depends not only on maintenance drug treatment, but also on family intervention, tackling comorbid drug misuse and the use of targeted individual psychological techniques such as monitoring of early signs and compliance therapy.

Strictly speaking, randomised controlled trials are still needed to confirm the effectiveness of early detection and intervention services. However, the testimony of patients and families, non-randomised evaluations of services such as those provided by the Early Psychosis Prevention and Intervention Centre service and obvious validity or common sense supports their wider introduction.

A first-episode service has potential benefits not only for patients and families: good organisational opportunities may also be taken by providers in setting up such a service. These include: emphasis on secondary prevention and normative models; prioritising of work issues; a built-in evaluative culture; strongly multi-disciplinary teamwork; built-in dual-diagnosis services; opportunities for cross-boundary initiatives including health, social services and the voluntary sector; inclusion of users and carers in planning and delivery of services; increased liaison with primary care; and opportunities for research in epidemiological cohorts. Evaluation at both the service and the individual level is crucial. There are snags that need to be thought through too, such as the danger of taking resources from hard-pressed general adult services; how a first-episode service relates logistically to the general adult service; and whether general clinicians excluded from such a service end up feeling further demoralised.