Published on 25 September 2026

Analysis · Schizophrenia · Psychopharmacology

Repurposed antihypertensives for negative symptoms in schizophrenia: what the meta-analysis can and cannot say

▤ Dossier Psychiatry and Clinical Neurosciences · 2025 · 79(12): 784-800 · Carnaval et al. DOI 10.1111/pcn.13892 PMID 40923586 Scientific 84 Editorial 68

Editorial collections A Study Under the Microscope

The essentials

This preregistered systematic review (PROSPERO CRD42022359199) pooled 12 double-blind, placebo-controlled randomised trials, 436 patients with schizophrenia or schizoaffective disorder, testing repurposed antihypertensives added to a stable antipsychotic: sodium nitroprusside, diuretics (diazoxide, bumetanide, spironolactone), telmisartan and clonidine. On the negative-symptom PANSS, the standardised mean difference is -0.37 (95% CI -0.59 to -0.15), with moderate heterogeneity (I² 43.8%); the effect remains significant after Holm correction (adjusted p 0.003). The prediction interval, however, runs from -0.82 to 0.09: in a new study, the true effect could be nil. The effect is larger in the nitroprusside subgroup, given as a 4-hour intravenous infusion (-0.67), which is no longer significant after multiplicity correction. GRADE certainty is moderate for the PANSS and low for adverse events. The authors themselves describe these results as preliminary. A preliminary lead, with no consequence for prescribing.

Context

Negative symptoms of schizophrenia, such as apathy, social withdrawal and poverty of speech, weigh heavily on daily functioning and remain poorly covered by antipsychotics. Hence the interest in repurposing drugs already marketed for other indications.

This meta-analysis asks a simple question: do antihypertensives, added to a stable antipsychotic, reduce these symptoms? The title speaks of antihypertensives as a block. What follows shows that the data do not establish a class effect.

The mechanism

The four categories pooled here rest on distinct mechanisms. Nitroprusside is a nitric oxide donor, telmisartan is an AT1 receptor antagonist that also has PPAR-γ agonist activity, and clonidine is an α2-adrenergic receptor agonist. Among the diuretics, bumetanide inhibits the NKCC1 cotransporter, spironolactone is thought to act on the NRG1-ERBB4 pathway, and diazoxide is a potassium channel opener. The authors put forward the hypothesis of improved cerebral perfusion common to several classes, which they describe as speculative: no included trial explored these pathways. The review measured a clinical effect on rating scales, not a mechanism.

The study at a glance

Item
Population
Adults with schizophrenia or schizoaffective disorder, on a stable antipsychotic (negative symptoms were not among the inclusion criteria)
Intervention
Repurposed antihypertensives as add-on: intravenous sodium nitroprusside (6 trials), diuretics (4: diazoxide, bumetanide or spironolactone), telmisartan (1), clonidine (1)
Comparator
Placebo, double-blind
Outcomes
Negative PANSS (primary), positive, general and total PANSS; rate of adverse events; exploratory: vital signs under nitroprusside
Design
Systematic review and meta-analysis of randomised trials, protocol registered in PROSPERO (CRD42022359199); PubMed, Web of Science and Scopus, English-language trials published since 2000, last search on 31 March 2025; 12 trials (13 comparisons), 436 participants; weighted mean follow-up of 8.1 weeks

Quality check

Item	Verdict
Protocol registration	Preregistered
FindingProtocol filed in PROSPERO.
Risk of bias of the trials	Some concerns
FindingAssessed with RoB 2 by two reviewers; 5 of 12 trials at low risk, 6 with some concerns, 1 at high risk. The search was limited to English-language publications since 2000, which the authors acknowledge as a possible source of bias.
Certainty of evidence	Explicit GRADE
FindingModerate for the PANSS (one level downgraded for risk of bias), low for adverse events (one level for risk of bias, one for imprecision).
Publication bias	Tested, except total PANSS
FindingEgger’s test and funnel plots for the subscales; test not possible for the total PANSS (too few studies). No publication bias detected, with slight visual asymmetries for the negative and general scales; the authors note that the power of the test is limited.
Multiplicity and robustness	Rigorous
FindingPrediction intervals, Holm correction across PANSS domains, leave-one-out analyses, an analysis restricted to trials at low risk of bias, handling of crossover and multi-arm trials.
Funding and conflicts of interest	No conflict declared
FindingPublic grant cited (FEDER/Ministerio de Ciencia, Innovación y Universidades, Agencia Estatal de Investigación, PID2023-147425OB-I00); no other source of funding is mentioned. Authors’ statement: no conflicts of interest.

Results

-0.37

Standardised mean difference on the negative PANSS, 12 trials, 436 patients (95% CI -0.59 to -0.15). Prediction interval -0.82 to 0.09.

Item	Value
Negative PANSS	-0.37
Finding95% CI -0.59 to -0.15; I² 43.8%; prediction interval -0.82 to 0.09; adjusted p (Holm) 0.003.
Positive, general, total PANSS	-0.29; -0.28; -0.44
FindingPooled estimates on the three other domains of the scale (11, 10 and 8 trials); they remain significant after Holm (adjusted p 0.014, 0.013 and 0.001). The authors describe the positive-symptom effect as marginal.
Nitroprusside subgroup	-0.67
Finding95% CI -1.26 to -0.07; I² 62.2%. No result in the subgroup remains significant after Holm correction. The Hallak trial (10 patients per arm) shows a markedly larger effect than the others (-2.61); a PEB recalculation from figure 7a, not published by the authors, gives about -0.38 without this trial.
Diuretics subgroup	-0.26
FindingNot significant (95% CI -0.64 to 0.11). On the total PANSS, -0.38 at nominal level (p = 0.027), not retained after Holm (p = 0.107). The authors consider that some diuretics may deserve study.
Telmisartan and clonidine	-0.19; -0.24
FindingNegative PANSS, one trial each (95% CI -0.79 to 0.41 and -0.94 to 0.45). No pooled estimate: excluded from the subgroup analyses.
Trials at low risk of bias	-0.77
FindingNegative PANSS in the analysis restricted to trials at low risk of bias: 95% CI -1.49 to -0.04; I² 70.9%. The effect remains in favour of treatment on all PANSS domains; this analysis is presented in a supplementary figure, not consulted.
Overall adverse events	IRR 0.82
FindingNot significant (p = 0.277); the 95% CI is reported as 0.56 to 1.20 in the figure and the GRADE table, and as 0.57 to 1.18 in the text. Under nitroprusside, IRR 0.58 (0.41 to 0.82), adjusted p 0.006, described as robust by the authors, with one caveat: exposure is assumed equal between a single infusion and treatments lasting several weeks. The telmisartan trial reports more adverse events on treatment (IRR 2.25, CI 1.10 to 4.62).
Vital signs under nitroprusside	No difference
FindingSystolic and diastolic blood pressure and heart rate showed no reliable difference between groups (5 trials), with high heterogeneity (I² 93.6%, 93.6% and 73.0%); exploratory analysis.

Critical appraisal

Item	Verdict
Methodology of the review	Solid, with limits
FindingPreregistration, RoB 2, explicit GRADE, prediction intervals, Holm correction, leave-one-out analyses, handling of crossover and multi-arm trials. Limits declared: means and standard deviations estimated from medians and ranges for some trials, a correlation of 0.5 assumed for change scores, data extracted from graphs, equal-exposure assumption for the IRRs, two trials excluded because of data presentation problems.
Pooling of mechanisms	Major reservation
FindingNitroprusside, diuretics (three molecules with distinct mechanisms), telmisartan and clonidine are grouped under the label of antihypertensives. The authors acknowledge the heterogeneity of the classes and conclude that some antihypertensives, above all nitroprusside, may have a benefit. A class effect is not demonstrated.
Prediction interval	Major reservation
FindingThe 95% CI excludes zero, but the prediction interval of -0.82 to 0.09 crosses it: in a new study, the true effect could be nil. The authors report this interval without commenting on it in the text. For the general and total domains, the prediction interval does not contain zero (-0.51 to -0.05 and -0.70 to -0.17).
Signal and multiplicity	Qualify
FindingThe effect is larger with nitroprusside, whose subgroup is no longer significant after Holm and whose estimate depends in part on one trial of 20 patients in total. The overall effects on the four PANSS domains remain significant after Holm. Nitroprusside is given as a 4-hour intravenous infusion; its place in routine practice has not been studied.
Small samples	Qualify
Finding436 patients in all, spread over four categories and six molecules, with short follow-up (weighted mean 8.1 weeks). The subgroups rest on 5 to 6 comparisons: numbers per molecule are limited, as the authors acknowledge.

Level of evidence

Scientific84/100

Editorial68/100

The level of evidence is that of a meta-analysis of randomised trials, conducted with a rigorous method on several points (preregistration, RoB 2, GRADE, multiplicity correction). Confidence is high in the quality of the synthesis; funding is public and no conflict of interest is declared.

It is markedly more reserved about what the synthesis allows one to conclude: the pooling of distinct mechanisms, a prediction interval for the negative PANSS that crosses zero, a nitroprusside subgroup effect that does not withstand multiplicity correction, and a modest total sample. A pooled average result is not a demonstrated effect.

The colleague test

What an experienced colleague would say if you presented this study in two minutes, between two consultations.

“An antihypertensive for negative symptoms? The effect is mostly seen with intravenous nitroprusside, and it is no longer significant in that subgroup once you correct for multiplicity. Interesting signal, preliminary evidence, nothing to prescribe.”

Translation for practice: follow this lead with interest, do not apply it, and explain to a patient or a colleague that an elegant average figure is not a treatment.

What you can do with this

What you can understand: an average effect of -0.37 on the negative PANSS, obtained by adding together very different molecules, does not tell you which one acts, or whether any of them does.
What you can tell a patient or a relative: this line of research exists, it is preliminary, and no antihypertensive has an established place in the treatment of negative symptoms.
What you can watch: ongoing or upcoming trials of a specific molecule, with a sufficient sample, rather than of a supposed class.
What you can teach: the difference between the confidence interval, which locates the average effect, and the prediction interval, which says what to expect from a new study.
Legal framework (France): the use of a drug outside its marketing authorisation is regulated in France by article L. 5121-12-1 of the Public Health Code, which makes it conditional on the absence of a suitable authorised alternative.

Frequently asked questions

Do antihypertensives improve negative symptoms in schizophrenia?

Not in a demonstrated way. The meta-analysis finds a modest average improvement on the negative PANSS, but it rests on very different molecules, with a larger effect for intravenous nitroprusside, and the prediction interval crosses zero. The overall effects remain significant after multiplicity correction; the authors conclude that the results are preliminary.

Why does the prediction interval change how the result reads?

The confidence interval locates the average effect of the included studies. The prediction interval gives the range in which the effect of a new study would fall. Here, on the negative PANSS, the first excludes zero and the second includes it: the uncertainty about a reproducible effect is real.

Can sodium nitroprusside be used in practice?

No established place. In the trials, it is given at 0.5 µg/kg/min for 4 hours intravenously, as a single or repeated infusion, in inpatients or outpatients; this route has not been evaluated in routine practice. Its effect in the subgroup where the effect is largest is also not significant after multiplicity correction.

Do spironolactone, telmisartan and clonidine have a place?

No established place. Telmisartan and clonidine are each represented by a single trial, and diuretics have a non-significant effect on the negative PANSS in the review; their signal on the total PANSS does not survive multiplicity correction.

Should we speak of a class effect?

No. The molecules pooled have distinct mechanisms, the hypothesis of a common mechanism (cerebral perfusion) remains speculative according to the authors, and the review does not allow one to speak of a class effect.

Annotated bibliography

Source study. Carnaval T, Bui M, Villoria J, Rodríguez D, Menchon JM, Ciruela F, Videla S. Repurposed antihypertensive drugs for negative symptoms in schizophrenia: A systematic review and meta-analysis. Psychiatry and Clinical Neurosciences, 2025, 79(12): 784-800. DOI: https://doi.org/10.1111/pcn.13892. PMID: 40923586. Preregistered systematic review, 12 trials, 436 participants. Funding: FEDER/Ministerio de Ciencia, Innovación y Universidades, Agencia Estatal de Investigación grant (PID2023-147425OB-I00). Conflicts of interest: “The authors have no competing interests to disclose.” No supplementary material came with the document consulted: the supplementary figures and tables of this publication were not consulted, and no data found only there is reported here.

Editorial collections

A Study Under the Microscope