Schizophrenia in adults according to NICE: the treatment decision tree, from first episode to clozapine
What the guidelines in force recommend for schizophrenia in adults. Time runs up the trunk, choices branch off it, and the canopy carries what applies at every step. Hover a node: the exact wording appears on the right, and the tree stays whole.
Intended readers. Health professionals. This tool replaces neither clinical judgement, nor the guidelines in force, nor the summary of product characteristics. It contains no dosages.
Read this before the tree. You are about to see two different things.
The strength of the recommendation answers one question: should it be done, and how firmly. That is what colours the tree by default. It is read in the wording of the guideline.
The certainty of the evidence answers a quite different question: how well do we know what this action produces. It is read in the studies, not in the guideline.
Neither follows from the other. A recommendation can be strong and rest on weak evidence, and that is not necessarily a mistake. This is why we never display them together.
This tree reads from the bottom up. The entry point sits at the very bottom, the outcomes and what to do at the top, and the step numbers climb with you.
Colour: the verb used
- Recommended. The text says “offer”.
- Consider. The text says “consider”.
- Advised against when the prohibition is plain, not routinely when the text attaches an exception to its refusal.
- Situation. A moment in the pathway, not a recommendation.
Shape: the nature of the action
- Drug or formulation.
- Psychological or family intervention.
- Investigation or check.
- Time interval or monitoring.
- Diamond. Clinical situation or branch point.
In the tree, each icon takes the colour of its node. The nature of the action never drives the colour.
Line: what the source does not say
- Dashed. An inferred step that no recommendation states. We draw it this way rather than pretend.
- Amber trunk. The chronology, from first episode upwards.
- Twigs. They rise towards the canopy, which depends on no step.
Entry to the pathway
Your pathway
- Click a card in the tree: the path leading to it from the entry point is listed here.
Nothing leaves your browser. What you type here is neither sent nor stored, on this site or anywhere else, and it is cleared when you close or reload the page. The printed sheet fits on one A4 page and keeps only the essentials of each step. Identify the patient by initials, never by name. Keeping the sheet in the medical record is your responsibility and falls under medical confidentiality.
The tree in full
What follows is not a copy of the drawing, it is its source. Each node is written once, here, with its coordinates, its parent, its colour and its badge carried in attributes. The script reads this text to draw the tree; it does not build it. There is therefore no second copy that could diverge.
The nodes of the tree
Person at increased risk of psychosis, without a defined episode
A side entry. This node carries the least known prohibition in the guideline.
Entry point.
What the guideline says
“ offer individual cognitive behavioural therapy (CBT) with or without family intervention ”CG178 1.2.3.1 [new 2014]
“ Do not offer antipsychotic medication: to people considered to be at increased risk of developing psychosis or with the aim of decreasing the risk of or preventing psychosis ”CG178 1.2.3.2 [new 2014]
“ monitor the person regularly for changes in symptoms and functioning for up to 3 years using a structured and validated assessment tool ”CG178 1.2.4.1 [new 2014]
What it does not say
The guideline sets no risk threshold, names no assessment tool, and gives no minimum duration of CBT specific to this situation.
The French regulatory frame
No French guideline covers this situation. The Haute Autorité de santé has begun the work: a scoping note on the detection and management of first episode psychosis and of the risk of progression to psychosis, adopted by the Board on 9 July 2025, approval expected in the first or second quarter of 2027.
What it rests on
No one has graded it. No source grades this step.
No analysis in the PEB corpus addresses this question. The branch stays dashed.
First episode psychosis
Entry point of the pathway.
Entry point.
Next: step 2, Oral antipsychotic, together with psychological interventions; step 2a, The person wants psychological treatment alone; step 2b, Do not start in primary care without psychiatric advice.
What the guideline says
“ Early intervention in psychosis services should be accessible to all people with a first episode or first presentation of psychosis, irrespective of the person’s age or the duration of untreated psychosis ”CG178 1.3.1.1 [new 2014]
“ Consider extending the availability of early intervention in psychosis services beyond 3 years if the person has not made a stable recovery from psychosis or schizophrenia ”CG178 1.3.1.4 [new 2014]
What it does not say
The guideline does not define first episode by operational criteria. It sets no duration of untreated psychosis beyond which management would change.
What it rests on
No one has graded it. No source grades this step.
None of the nine PEB analyses on schizophrenia addresses first episode psychosis in adults. That is a gap in the corpus, and it is flagged rather than hidden.
Oral antipsychotic, together with psychological interventions
The antipsychotic on its own is not what is recommended.
Comes after step 1, First episode psychosis.
Next: step 3, Choosing the drug.
What the guideline says
“ For people with first episode psychosis offer: oral antipsychotic medication in conjunction with psychological interventions (family intervention and individual CBT) ”CG178 1.3.4.1 [new 2014]
What it does not say
No order of preference between the two components. No indication of what to do if one of them is not available locally.
What it rests on
No one has graded it. No source grades this step.
No PEB analysis attached to this step.
The person wants psychological treatment alone
A common consultation situation, missing from most decision trees.
Comes after step 1, First episode psychosis.
What the guideline says
“ Advise people who want to try psychological interventions alone that these are more effective when delivered in conjunction with antipsychotic medication. ”CG178 1.3.4.2 [new 2014]
What it does not say
The guideline does not say what to do if, once the interval has elapsed, the person still refuses.
What it rests on
No one has graded it. No source grades this step.
The one month or less review interval is graded by no source.
Do not start in primary care without psychiatric advice
Comes after step 1, First episode psychosis.
What the guideline says
“ Do not start antipsychotic medication for a first presentation of sustained psychotic symptoms in primary care unless it is done in consultation with a consultant psychiatrist ”CG178 1.3.2.1 [2009 ; amended 2014]
The French regulatory frame
In France, the ALD 23 schedule of July 2025 places the general practitioner and the psychiatrist at the initial assessment of every patient. The document is medico-administrative and prescribes no strategy.
What it rests on
No one has graded it. No source grades this step.
No grading. A recommendation whose evidence has not been re-examined since 2009, according to NICE’s own labelling.
Choosing the drug
A shared decision. And a fact most decision trees conceal.
Comes after step 2, Oral antipsychotic, together with psychological interventions.
Next: step 4, Baseline investigations before starting.
What the guideline says
“ The choice of antipsychotic medication should be made by the service user and healthcare professional together, taking into account the views of the carer if the service user agrees. ”CG178 1.3.5.1 [2009 ; amended 2014]
What it does not say
NICE names no first line drug. Nor does the text rank the generations. The only drugs named anywhere in the guideline are clozapine, olanzapine and chlorpromazine, never as an initial choice. That is a position, not an omission.The French regulatory frame
The choice is made among the drugs licensed for schizophrenia in France.
What it rests on
No one has graded it. No source grades this step.
No source grades the first line choice, for the simple reason that none settles it.
Baseline investigations before starting
Comes after step 3, Choosing the drug.
Next: step 5, The treatment trial, at optimum dosage, 4 to 6 weeks; step 5a, Two prohibitions at initiation.
What the guideline says
“ Before starting antipsychotic medication, undertake and record the following baseline investigations ”CG178 1.3.6.1 [2014, amendé 2021]
“ Before starting antipsychotic medication, offer the person an electrocardiogram (ECG) if ”CG178 1.3.6.2 [2009]
What it does not say
No numerical threshold. No body mass index, no waist circumference, no blood glucose, no corrected QT. The guideline says what to measure, never from what value to act.
The French regulatory frame
The ALD 23 schedule of July 2025 added three investigations to the list of covered procedures: serum magnesium, high sensitivity CRP and troponin. The last two carry the note “baseline and follow-up for clozapine”.
What it rests on
No one has graded it. No source grades this step.
No grading, either in the guideline or in the corpus.
The treatment trial, at optimum dosage, 4 to 6 weeks
The threshold that structures every French language decision tree.
Comes after step 4, Baseline investigations before starting.
Next: step 6, At 4 to 6 weeks, is the response adequate?
What the guideline says
“ Carry out a trial of the medication at optimum dosage for 4–6 weeks ”CG178 1.3.6.3 [2009 ; amended 2014]
“ At the start of treatment give a dose at the lower end of the licensed range and slowly titrate upwards within the dose range given in the British national formulary (BNF) or SPC ”CG178 1.3.6.3 [2009 ; amendé 2014]
What it does not say
This recommendation carries the [2009] label. NICE defines it thus: “the evidence has not been reviewed since 2009”. The 4 to 6 week threshold has not been re-examined for seventeen years.What it rests on
No one has graded it. No source grades this step.
This threshold is graded by no source. Furukawa’s network meta-analysis defines its population by “at least one adequate four week antipsychotic trial”, which documents the use of the threshold, not its validity.
Two prohibitions at initiation
The most instructive node. Look at it in both views.
Comes after step 4, Baseline investigations before starting.
What the guideline says
“ Do not use a loading dose of antipsychotic medication (often referred to as ‘rapid neuroleptisation’) ”CG178 1.3.6.9 [2009]
“ Do not initiate regular combined antipsychotic medication, except for short periods (for example, when changing medication) ”CG178 1.3.6.10 [2009]
What it rests on
Very low certainty. Level assigned by PEB from the analyses cited.
On combining two antipsychotics in non-responders, the network meta-analysis by Furukawa and colleagues, eClinicalMedicine 2026, 18 trials and 575 patients, gives a standardised mean difference of −0.25, interval −0.46 to −0.04, with very low certainty, and an odds ratio of 1.93, interval 1.26 to 3.00, for adverse effects.
In real life, the French claims-data cohort of Llorca and colleagues, Comprehensive Psychiatry 2026, 234,959 patients, finds exposure to polypharmacy in 71.1% of the cohort. Oxford level of evidence 2b.
A firm negative recommendation, a majority practice, and very low evidence on both sides. Contradicting this recommendation in practice rests on nothing better than the recommendation itself.
At 4 to 6 weeks, is the response adequate?
Two paths from here.
Comes after step 5, The treatment trial, at optimum dosage, 4 to 6 weeks.
Next: step 7, Maintenance, monitoring and annual review; step 7a, The four checks, before changing anything.
No direct wording
No recommendation states this step in its own words.
What it does not say
NICE does not define response. No scale, no percentage reduction, no operational criterion. The most used node of every decision tree is the one the guideline leaves emptiest.
What it rests on
No one has graded it. No source grades this step.
Nothing to grade: the guideline defines neither the response, nor its threshold, nor the scale that measures it.
Maintenance, monitoring and annual review
Left hand path.
Comes after step 6, At 4 to 6 weeks, is the response adequate?
Next: step 8b, Long-acting injectable formulation; step 8, Intermittent dosing strategy.
What the guideline says
“ Monitor and record the following regularly and systematically throughout treatment, but especially during titration ”CG178 1.3.6.4 [2014, amendé 2021]
“ The secondary care team should maintain responsibility for monitoring service users’ physical health and the effects of antipsychotic medication for at least the first 12 months or until the person’s condition has stabilised ”CG178 1.3.6.5 [new 2014]
“ Review antipsychotic medication annually, including observed benefits and any side effects ”CG178 1.5.1.3 [new 2014]
“ Inform the service user that there is a high risk of relapse if they stop medication in the next 1–2 years ”CG178 1.4.6.3 [2009]
“ After withdrawal from antipsychotic medication, continue monitoring for signs and symptoms of relapse for at least 2 years ”CG178 1.4.6.5 [2009]
What it does not say
No total duration of maintenance treatment after a first episode. The guideline warns of the relapse risk on stopping, it never says how long to treat.
The French regulatory frame
Handover to the family doctor then takes place at least once a year, under recommendation 1.5.3.2. In France, cardiometabolic monitoring under antipsychotics is covered by position statements from the national medicines agency.
What it rests on
No one has graded it. No source grades this step.
The monitoring intervals are graded by no source. They are follow-up conventions, not trial results.
Go and look for yourself
Long-acting injectable formulation
Wording currently under revision.
Comes after step 7, Maintenance, monitoring and annual review.
What the guideline says
“ Consider offering depot/long-acting injectable antipsychotic medication to people with psychosis or schizophrenia ”CG178 1.5.5.3 [2009]
“ ”CG178 1.5.6.1 [2009, amendé 2022]
What it does not say
No injection interval, no dosage, no drug named. And no recommendation for an injectable at first episode: these recommendations sit in the section on recovery, that is, after an acute episode.
The French regulatory frame
An open point, worth knowing before prescribing. NICE’s surveillance report of 9 November 2017 concludes that an amendment would be made to this recommendation, removing the initial test dose, on the grounds that some recent injectable formulations require prior stabilisation on the equivalent oral form, which would make the test dose pointless under their licence. The version we consulted still carries the old wording, and the text’s update log records no change in 2017. We therefore cannot establish that this amendment was applied. In France, initiation is governed by each drug’s summary of product characteristics.What it rests on
No one has graded it. No source grades this step.
None of the nine PEB analyses on schizophrenia covers long-acting injectables. We will not attach to this branch the analysis on adherence and mortality: it states explicitly that it does not assess adherence patterns by type of antipsychotic.
Saying so is more useful than hanging a reference that does not support the branch.
Intermittent dosing strategy
A qualified negative, with an exception written into the text.
Comes after step 7, Maintenance, monitoring and annual review.
What the guideline says
“ Do not use targeted, intermittent dosage maintenance strategies routinely. However, consider them for people who are unwilling to accept a continuous maintenance regimen ”CG178 1.5.5.2 [2009]
What it does not say
The recommendation is negative with an exception, which is unusual. The intermittent strategy is defined in the text as using an antipsychotic only during periods of incipient relapse or exacerbation, rather than continuously.
What it rests on
No one has graded it. No source grades this step.
Not graded. The 2017 surveillance report notes that new evidence was found on intermittent techniques, and judged it insufficient to add a recommendation.
The four checks, before changing anything
Right hand path. To be done before touching the treatment.
Comes after step 6, At 4 to 6 weeks, is the response adequate?
Next: step 8a, Second antipsychotic.
What the guideline says
“ For people with schizophrenia whose illness has not responded adequately to pharmacological or psychological treatment: review the diagnosis ”CG178 1.5.7.1 [2009]
What it does not say
The third check is about psychotherapy, not medication. It asks you to verify that psychological interventions were in fact offered, and to offer another one where relevant. It is almost always missing from French language decision trees, which keep only adherence and substance use.The French regulatory frame
The ALD 23 schedule of July 2025 lists plasma level measurement of psychotropics and urine drug screening among covered procedures, on an “as needed” basis.
What it rests on
No one has graded it. No source grades this step.
Not graded.
Second antipsychotic
No recommendation states this step.
Comes after step 7a, The four checks, before changing anything.
Next: step 9, Clozapine.
No direct wording
No recommendation states this step in its own words.
What it does not say
Nowhere does NICE recommend switching antipsychotic after a first failure. The very existence of a second line is inferred from a phrase slipped into the next recommendation. Drawing a solid arrow here would be an extrapolation.
What it rests on
Very low certainty. Level assigned by PEB from the analyses cited.
The network meta-analysis by Furukawa and colleagues, eClinicalMedicine 2026, 59 trials and 5,409 participants, classifies switching antipsychotic and increasing the dose both as inconclusive. At this step, the literature settles nothing.
Go and look for yourself
Clozapine
Comes after step 8a, Second antipsychotic.
Next: step 10, Inadequate response to clozapine at an optimised dose.
What the guideline says
“ Offer clozapine to people with schizophrenia whose illness has not responded adequately to treatment despite the sequential use of adequate doses of at least 2 different antipsychotic drugs ”CG178 1.5.7.2 [2009]
What it does not say
No dosage, no blood count schedule, no minimum duration for the two prior trials. Every practical safeguard around this drug comes from elsewhere.
The French regulatory frame
In France, clozapine is reserved for failure, either non-response or poor tolerability, of at least two antipsychotics including at least one second-generation drug. Its prescription is governed by neutrophil count monitoring defined by the national medicines agency. The ALD 23 schedule of July 2025 added high sensitivity CRP and troponin to the baseline assessment and follow-up of this drug.
What it rests on
Very low certainty. Level assigned by PEB from the analyses cited.
The individual participant data meta-analysis by Schneider-Thoma and colleagues, The Lancet Psychiatry 2025, 19 trials and 1,599 participants, gives a mean difference of −0.64, credible interval −3.97 to 2.63, on symptoms at six to eight weeks. GRADE confidence very low.
What this result does not say, and must be read before concluding. It concerns a short-term symptom scale. The outcomes that ground the recommendation are not measured in this analysis: rehospitalisation, suicidality, mortality and treatment retention are all recorded there as not reported. An absence of difference on a surrogate outcome at eight weeks is not proof of absence of effect, still less on outcomes that were never looked at.
NICE examined this question and published its reasoning. Its 2017 surveillance report analysed a 2016 network meta-analysis, 40 trials and 5,172 participants, selected precisely because it “may contradict current recommendations for the use of clozapine”. NICE’s conclusion, word for word: the overall population was not the one the recommendation targets, and “when this population was evaluated individually, no significant difference in antipsychotic efficacy was reported”. NICE nonetheless maintains the recommendation, “in combination with the evidence considered during guideline development”.
In other words, the body that recommends clozapine itself acknowledges that, in exactly the population targeted, no significant difference in antipsychotic efficacy was reported. It owns the judgement call, and writes it down. That is more honest than most decision trees, which reproduce this branch without ever discussing it.
Inadequate response to clozapine at an optimised dose
The end of the guideline.
Comes after step 9, Clozapine.
What the guideline says
“ Consider recommendation 1.5.7.1 (including measuring therapeutic drug levels) before adding a second antipsychotic to augment treatment with clozapine ”CG178 1.5.7.3 [2009]
What it does not say
No augmentation drug is named. The only criterion of choice given is negative: do not compound the common side effects of clozapine.
What it rests on
Low certainty. Level assigned by PEB from the analyses cited.
The network meta-analysis by Samara and colleagues, eClinicalMedicine 2025, 150 trials and 11,375 participants, places nine augmentation strategies ahead of clozapine alone on the primary outcome, with estimates ranging from −1.67 to −0.51. The strongest, duloxetine added to clozapine, rests on a single trial of 40 patients. A small-study effect is found in this network.
When clozapine is not possible, the meta-analysis by Carr and colleagues, Molecular Psychiatry 2025, 68 studies and 3,241 patients, maps nine augmentation classes. All are rated low or very low. None reaches moderate.
Go and look for yourself
- When clozapine is not possible, what is actually left?
- What is an effect size of minus 1.53 for duloxetine worth?
- The step number, in a pill at the corner of each card. It gives the rank along the path. Cards of the same rank on another branch are marked a, b, c.
- Pale blue glow. On hover or click, the steps that may follow light up in pale blue, and the path already travelled down to the entry lights up in amber.
What applies at every step
Recommendations that do not depend on position in the pathway. Canopy entries are summaries of the guideline, not word for word quotations.
Social skills
Social skills training
PEB summary
Not offered routinely as a specific intervention.
Where it is written
CG178 1.4.4.8
Physical activity
Combined healthy eating and physical activity programme
PEB summary
Offered by the mental health service, particularly for people taking antipsychotics.
Where it is written
CG178 1.1.3.1
Smoking cessation
Help to stop smoking
PEB summary
Offered even after unsuccessful attempts. Note the effect of reduced smoking on the metabolism of clozapine and olanzapine. The guideline explicitly rules out bupropion in people with psychosis, and asks that people taking bupropion or varenicline be warned of the increased risk of adverse neuropsychiatric symptoms, with regular monitoring, particularly in the first 2 to 3 weeks (1.1.2.4).
Where it is written
CG178 1.1.3.3
Individual CBT
Individual cognitive behavioural therapy
PEB summary
At least 16 planned sessions, delivered according to a treatment manual.
Where it is written
CG178 1.3.7.1
Family intervention
Family intervention
PEB summary
From 3 months to 1 year, at least 10 planned sessions.
Where it is written
CG178 1.3.7.2
Counselling and support
Counselling and supportive psychotherapy
PEB summary
Not offered routinely as specific interventions.
Where it is written
CG178 1.4.4.6
Adherence therapy
Adherence therapy
PEB summary
Not offered as a specific intervention.
Where it is written
CG178 1.4.4.7
Catatonia, and what the guideline says about it. It is addressed at a single point, recommendation 1.5.7.4, which refers to a separate document, technology appraisal TA59 of 2009: electroconvulsive therapy is there an option for rapid short-term treatment of severe symptoms of catatonia when other treatments have not worked or the condition is life-threatening. This tree does not make a node of it, because the detailed course of action belongs to that other document, which PEB has not read.
Our quotation regime. This tree quotes one sentence per node, the one that decides, with its recommendation number and date label. It reproduces no bullet list from the source text: those are summarised and announced as summaries. Twenty-six short quotations, none beyond thirty-five words. A reader who wants to check word for word follows the link to NICE. This choice follows from the fact that the NICE open content licence covers the United Kingdom only.
How this tree was built
The 26 quotations come from the NICE text itself, taken from the source and never from memory, with their recommendation number and date label. When the NICE site became reachable again, on 28 August 2026, every claim of absence was reopened rather than carried over: two of them were wrong, and this page has corrected them.
The detail of these checks, what could be verified and what could not, is set out on the page Comment PEB vérifie ce qu’elle publie.
Sources
The full critical appraisal of this guideline, with its AGREE II methodological assessment, is published separately: NICE CG178 guideline record.
- National Institute for Health and Care Excellence. Psychosis and schizophrenia in adults: prevention and management. NICE Clinical Guideline No. 178. London, March 2014. Replaces guideline CG82 of March 2009. Consulted in its NCBI Bookshelf reproduction, identifier NBK555203, ISBN 978-1-4731-0428-0. This version carries an August 2019 amendment on access to early intervention services.
- National Institute for Health and Care Excellence. Surveillance report 2017, Psychosis and schizophrenia in adults: prevention and management (2014) NICE guideline CG178. Published 9 November 2017. ISBN 978-1-4731-1383-1.
- Haute Autorité de santé. ALD 23, Schizophrénies. Actes et prestations affection de longue durée. Approved by the Board on 17 July 2025.
- Haute Autorité de santé. Repérage et prise en charge des personnes présentant un premier épisode psychotique ou un risque d’évolution vers une psychose. Scoping note. Adopted by the Board on 9 July 2025.
- Haute Autorité de santé. Diagnostic et évaluation des troubles schizophréniques. Scoping note. Adopted by the Board on 8 April 2026.
And across the Channel
This tree sets out what the British guideline says. It names no drug. The French regulatory frame, by contrast, names the twenty licensed drugs and says nothing about strategy. The two are read one after the other, and the gap between them is the point.
Open the French algorithmPsychiatry Evidence Base, evidence-based psychiatry, explained with rigour.