Published on 24 September 2026
Varenicline and bupropion for alcohol use disorder: the trial announces a combination, it demonstrates a monotherapy
The essentials
This Swedish phase 2, randomized, double-blind, four-arm, placebo-controlled trial compared, over thirteen weeks, varenicline combined with bupropion, varenicline alone, bupropion alone, and a double placebo, in 384 adults with moderate to severe alcohol use disorder. Against placebo, the combination reduced blood phosphatidylethanol (d=0.39, p=0.004) and the percentage of heavy drinking days (d=0.31, p=0.008). Varenicline alone did the same (d=0.30, p=0.005 and d=0.36, p=0.023). But the comparison that carried the trial’s hypothesis, the combination against varenicline alone, showed no difference (d=0.022, p=0.97 on the biomarker, d=0.027, p=0.76 on heavy drinking days), and the prespecified sequential testing hierarchy stopped exactly there. The bupropion-alone arm was tested after this stop: its effect sizes on the two co-primary endpoints range from 0.14 to 0.28, and nothing it produces retains confirmatory value. In other words, the demonstrated benefit is attributable to varenicline. Bupropion’s own contribution lies elsewhere, and it is secondary: it shortens the duration of varenicline-induced nausea. None of this transposes to France, where both molecules are indicated for smoking cessation, not for alcohol.
Context
Alcohol use disorder is one of the presenting complaints where the gap between how common the problem is and how much room pharmacological treatment actually occupies is widest. Authorized molecules for this indication exist and are effective, but their average effect size remains modest, and a large proportion of patients receive none of them. Any lead capable of shifting that level of efficacy deserves serious examination, including when it comes from molecules developed for a different dependence.
Varenicline and bupropion come from smoking cessation. Testing them in alcohol use disorder is not a matter of opportunistic repurposing: it follows an explicit pharmacological hypothesis, and the clinical overlap between alcohol and nicotine is strong, since 49 percent of this trial’s participants were daily nicotine users. The question the authors ask is not only whether varenicline acts on alcohol consumption, but whether adding bupropion amplifies that effect. This is the combination hypothesis, and it is what gives the trial its title.
Mechanism
What the trial tests, and what it does not measure
The working hypothesis is dopaminergic. Alcohol’s reinforcing effect is thought to pass largely through the mesolimbic dopaminergic system, whose activation is modulated by nicotinic acetylcholine receptors. Varenicline, a partial agonist at these receptors, is meant to attenuate that reinforcement. Bupropion, which acts on dopamine and noradrenaline reuptake, is meant to act on the same system through a different pathway, hence the idea of an additive effect from the combination.
The status of this hypothesis within the trial needs to be stated precisely. The outcome measures are all measures of consumption, abstinence, or craving. None directly documents the proposed mechanism: no imaging, no receptor occupancy, no central marker. The trial therefore tests a clinical consequence expected from the dopaminergic hypothesis, it does not test the hypothesis itself. A negative result on the combination does not refute the pharmacological model, and a positive result on varenicline does not prove it either.
What blood phosphatidylethanol measures
Blood phosphatidylethanol, or B-PEth, is an abnormal phospholipid formed in the red blood cell membrane only in the presence of ethanol. Its concentration reflects cumulative alcohol exposure over the preceding weeks, where blood alcohol level covers only a few hours and a drinking diary depends on what the patient reports. The authors claim to be the first randomized trial in alcohol use disorder to make it a primary endpoint, and this is probably the most durable methodological contribution of this work: on this endpoint, reporting bias disappears.
What it does not measure must nonetheless be named. B-PEth quantifies exposure, not disease. It says nothing about craving, nothing about functional, family, or occupational impact, nothing about somatic complications, nothing about mortality. It remains a surrogate marker of drinking behavior, objective but intermediate, not a hard clinical endpoint. A drop in B-PEth means the patient drank less over the period, which is already a great deal, but translating that drop into durable clinical benefit is an inference that this trial, lasting thirteen weeks with no follow-up after treatment stopped, does not document.
The study at a glance
| Population, intervention, comparator, outcomes | |
|---|---|
| Population | |
| 384 adults aged 25 to 70 years analyzed, moderate to severe alcohol use disorder (at least 4 of 11 DSM-5 criteria), B-PEth at least 0.5 µmol/L at screening, at least two heavy drinking days per week over the preceding three months. Recruitment through media advertisements. Numerous exclusions: abstinence between screening and randomization, treatment for withdrawal within 30 days, any treatment likely to affect consumption, psychiatric or somatic comorbidity likely to affect assessments or safety, a history of delirium tremens or withdrawal seizure within the past five years. Population 72 percent male, median age 57 years, 49 percent daily nicotine users. | |
| Intervention | |
| Varenicline 1 mg twice daily combined with extended-release bupropion 150 mg twice daily, after a one-week titration period, then 12 weeks at full dose with no dose adjustment. | |
| Comparators | |
| Three other arms, in a 1:1:1:1 allocation: placebo plus placebo, varenicline plus placebo, placebo plus bupropion. These two monotherapy arms are what gives the design its value: they make it possible to attribute the effect, which most combination trials do not allow. | |
| Outcome measures | |
| Two co-primary endpoints over weeks 2 to 11: blood phosphatidylethanol and the percentage of heavy drinking days, collected by timeline follow-back. Secondary endpoints, with no stated hierarchy among them: total consumption in grams per day, percentage of abstinent days, craving measured on a visual analogue scale. | |
| Design and sample size | |
| Phase 2 randomized controlled trial, double-blind, placebo-controlled, multicentre across four Swedish sites, duration 13 weeks, that is 91 days. 388 patients randomized, 384 analyzed in the modified intention-to-treat population: 97 on double placebo, 100 on varenicline plus bupropion, 96 on varenicline plus placebo, 91 on placebo plus bupropion. Per-protocol population, defined by at least 80 percent of treatment days taken, ranging from 61 to 69 patients per arm. Oxford CEBM level of evidence 1b. |
Quality control
| Point checked | Verdict |
|---|---|
| Randomization and allocation | Blocks of 8, managed by a third party |
| FindingThe sequence was generated and held by a third party, the dispensing pharmacist, who also distributed treatments to the sites. The code was broken only after database lock, and the statistical analysis was conducted on coded groups. An independent good clinical practice audit was carried out by a provider named in the article. | |
| Blinding maintained | Functional unblinding likely |
| FindingNausea occurred in 51 percent of patients on varenicline alone, that is 49 of 96, versus 11 percent on double placebo, that is 11 of 97, with a median duration of 45 days versus 10 days. An adverse effect this discriminating allows allocation to be inferred, which affects the self-reported endpoint. No blinding-integrity test is reported, either in the methods or in the appendix. | |
| Outcome measures | A blood biomarker as co-primary |
| FindingChoosing an objective co-primary endpoint protects the main result from reporting bias, which is uncommon in this field. The second co-primary remains self-reported, and it is the one exposed to unblinding. | |
| Sample size and power | Target reached |
| FindingPower was calculated a priori with an alpha risk of 0.025 and a power of 0.80, for a detectable effect of d=0.44 on the biomarker and d=0.43 on heavy drinking days. The target of 360 patients in the modified intention-to-treat population was reached. This detectable-effect value is worth keeping in mind when reading the negative results. | |
| Missing data | No imputation |
| FindingMeans are calculated on available data, with no imputation, which the authors state explicitly. Arm-by-arm dropout details appear in the participant flow diagram, which we were not able to read. What can be established concerns compliance: the per-protocol population retains 64 of 100 patients in the combined arm, versus 69 of 97 on double placebo and 68 of 96 on varenicline alone. The arm carrying the most medication is the one that retains the fewest patients. | |
| Analysis plan | Prespecified sequential hierarchy |
| FindingThe two co-primary endpoints are treated as two independent hypotheses, each at an alpha risk of 0.025 in the sample size calculation, with no multiplicity adjustment between them or between comparisons, which are instead hierarchically ordered. A Holm-Bonferroni correction is applied to secondary endpoints. Dual preregistration and a published protocol; the statistical analysis plan and the data management plan are not published, they are available on request. The design is transparent, and that is precisely what makes visible the point where it stops. | |
| Presentation of results | Per-protocol foregrounded |
| FindingThe per-protocol analysis, restricted to the 61 to 69 most compliant patients per arm, yields numerically larger effect sizes for the combination, and it is this analysis that is foregrounded in the conclusion. | |
| Safety | Eleven serious events, evenly distributed |
| FindingEleven serious adverse events in total, distributed evenly across arms, none considered to have a high probability of being treatment-related. A thirteen-week phase 2 trial in 384 patients cannot settle the question of rare events. | |
Results
| Comparison | Result |
|---|---|
| Varenicline plus bupropion versus placebo, co-primary endpoints | B-PEth d=0.39, p=0.004; heavy drinking days d=0.31, p=0.008 |
| ReadingThe combination outperformed double placebo on both co-primary endpoints, in the modified intention-to-treat population. This is the result foregrounded by the title and the abstract. | |
| Varenicline alone versus placebo, co-primary endpoints | B-PEth d=0.30, p=0.005; heavy drinking days d=0.36, p=0.023 |
| ReadingMonotherapy also outperformed double placebo on both endpoints, with values of the same order. On heavy drinking days, the effect size for varenicline alone is numerically higher than that of the combination. | |
| Combination versus varenicline alone | B-PEth d=0.022, p=0.97; heavy drinking days d=0.027, p=0.76 |
| ReadingNo difference was detected on either co-primary endpoint. On the biomarker, the difference estimate is -0.002, with a 95 percent confidence interval of -0.123 to 0.119. The prespecified sequential hierarchy stops at this comparison: strictly speaking, no subsequent test retains confirmatory evidentiary status. | |
| Bupropion alone versus placebo, co-primary endpoints | Outside the hierarchy; effect sizes 0.14 to 0.28 |
| ReadingThis comparison falls after the hierarchy stopped, so it carries no confirmatory value. It is not, however, empty: on heavy drinking days, the nominal p value is 0.04 in both the modified intention-to-treat and per-protocol populations, and the appendix reports, without multiplicity correction, a reduction in total consumption (d=0.30, p=0.041 in the modified intention-to-treat population) and an increase in the proportion of abstinent days (d=0.41, p=0.024). Read this alongside the trial’s detectable effect, on the order of d=0.43 to 0.44: a smaller true effect could not have been detected by this design. | |
| Total consumption in grams per day, after multiplicity correction | Significant for the combination: d=0.44 in the modified intention-to-treat population, d=0.47 per protocol |
| ReadingThis is the only secondary endpoint that survives multiplicity correction in the modified intention-to-treat population, and it does so for the combination only. Per protocol, craving also survives, but for varenicline alone (d=0.43, p=0.005). These results fall after the hierarchy stopped, which gives them exploratory rather than confirmatory value. | |
| Per-protocol analysis, 61 to 69 patients per arm | Combination d=0.43 and 0.41; varenicline alone d=0.29 and 0.34 |
| ReadingThe combination’s effect sizes here are numerically larger than varenicline alone’s. These two sets of values were never directly compared with a statistical test. Comparing two effect sizes each obtained against placebo does not substitute for the direct comparison, which was also carried out per protocol, and which is likewise negative there (p=0.78 on the biomarker, p=0.63 on heavy drinking days). | |
| Nausea | 51 percent on varenicline alone versus 11 percent on double placebo; median duration 45 days, 10 days on double placebo, 16.5 days when bupropion is combined |
| ReadingThis is bupropion’s own contribution in this trial: it does not strengthen varenicline’s effect on consumption, it shortens one of its notable adverse effects. This is a tolerability result, located among the secondary endpoints, not a primary outcome. | |
| Serious adverse events | Eleven in total, evenly distributed, none judged to have a high probability of being treatment-related |
| ReadingNo particular safety signal over the course of the trial. The sample size and the thirteen-week duration do not allow any conclusion about rare or late-emerging events. | |
One reading note, a favorable one this time. The results tables report, for each comparison, the difference estimate and its 95 percent confidence interval. The negative comparisons can therefore be bounded, which is rare and welcome: on the biomarker, the difference between the combination and varenicline alone ranges from -0.123 to 0.119, a tight interval centered on zero. No confidence interval is given, however, around the standardized effect sizes themselves, which remain point estimates.
Critical appraisal
| Domain | Risk of bias |
|---|---|
| Domain 1, randomization process | Low risk |
| FindingSequence in blocks of 8 generated and held by a third party, allocation concealed, blinding maintained until database lock. | |
| Domain 2, deviations from intended intervention | Some concerns |
| FindingThe very marked asymmetry in nausea between arms makes functional unblinding plausible, though this cannot be confirmed retrospectively, since no blinding-integrity test was reported. | |
| Domain 3, missing outcome data | Some concerns |
| FindingNo imputation of missing data, and uneven compliance across arms, lower in the combined arm. The much-reduced per-protocol analysis is, moreover, foregrounded in the conclusion, which exposes the result to adherence bias. | |
| Domain 4, measurement of the outcome | Low risk |
| FindingThe first co-primary is a blood assay, insensitive to reporting bias. The second is self-reported and therefore exposed to the unblinding discussed in domain 2. The overall risk-of-bias assessment places the concerns in domains 2 and 3, and we do not tighten it here. | |
| Domain 5, selection of the reported result | Low risk |
| FindingDual preregistration and a published protocol. The authors explicitly report the non-significance of the comparison between the combination and varenicline alone, as well as the stop in the hierarchy. | |
| Overall risk of bias, RoB 2 tool | Some concerns |
| FindingThe overall level retained is some concerns, driven by domains 2 and 3. This is not a high risk, and the trial remains usable: it is a methodologically well-built trial with two identified points of fragility. | |
| Fit between the claim and the evidence | Title and conclusion ahead of the data |
| FindingThe title, abstract, and conclusion foreground the combination, whereas the only direct comparison concerning it is negative and the hierarchy stops there. The implicit argument rests on numerically larger effect sizes per protocol, which were never statistically compared against each other. The discussion also compares the observed effect sizes to those reported for naltrexone and acamprosate in a meta-analysis (d=0.2), an indirect comparison across different trials, not tested here. | |
| Independence and conflicts of interest | Declared ownership conflict |
| FindingFive of eleven authors are co-owners of the company Sobrera Pharma AB, including the first author and the last author, who are its founders. Other authors declare honoraria or research funding from several pharmaceutical companies. These ties are disclosed. The sponsor remains academic, the primary funding public, the analysis was conducted blinded on coded groups, and an independent audit took place; the authors state that Sobrera Pharma neither produced nor supplied the study drugs and took part in neither the statistical analysis nor the data interpretation. This is nonetheless an ownership conflict, the strongest degree of conflict of interest, and it is something the reader should keep in mind when reading the foregrounding of the combination in the title and conclusion. We presume no intent, we flag a configuration. | |
| External validity | Selected population |
| FindingRecruitment through media advertisements and numerous exclusion criteria, in particular any psychiatric or somatic comorbidity likely to affect assessments or safety, and any history of delirium tremens or withdrawal seizure within the past five years. The result is a socially stable and medically healthy population, median age 57 years, fairly far removed from the usual addiction-medicine caseload, where comorbidities are the rule. | |
Level of evidence
Confidence is high on one point, and it is far from minor: in this population and over thirteen weeks, varenicline reduces alcohol consumption compared with placebo, and this result holds on both co-primary endpoints, one of them a blood assay insensitive to reporting bias. This is demonstrated, in the sense of a single phase 2 randomized trial with a risk of bias rated some concerns. What is demonstrated concerns monotherapy, whatever the trial’s title says.
Confidence is low, or absent, on three points, and the same symmetry applies to each of them. First, the combination showed no superiority over varenicline alone, which does not establish their equivalence: the trial was not designed as an equivalence or non-inferiority trial, and no clinically acceptable margin was set a priori. The confidence interval of the difference is, however, published, and it is tight around zero on the biomarker, -0.123 to 0.119, which makes a substantial benefit from adding bupropion on this endpoint unlikely. What can be said is therefore simple: nothing in this trial supports adding bupropion to improve the effect on consumption. Second, the bupropion-alone arm reads within the same limits, compounded by its position in the hierarchy: tested after the stop, it produces no confirmatory result, its effect sizes on the co-primaries range from 0.14 to 0.28 while the design was powered to detect an effect on the order of d=0.43 to 0.44, and a few nominal p values below 0.05 appear that cannot be interpreted. Not testable does not mean ineffective, for bupropion any more than for the combination. Third, the durability of the effect is not documented: the trial stops at thirteen weeks, with no follow-up after treatment ended.
One distinction of status remains to be kept clear. That varenicline reduces consumption in this selected population is demonstrated. That its effect is larger than that of treatments currently authorized for this indication is suggested, on the basis of an indirect comparison across different trials, never tested here head to head. That the nicotinic and dopaminergic pathway is a relevant target in alcohol use disorder belongs to the authors’ working hypothesis, consistent with their results, but not something this trial measures directly.
The colleague test
What an experienced colleague might say about this study in two minutes, between two consultations.
“Honestly, a well-conducted Swedish trial, and a blood biomarker as the primary endpoint, that’s interesting. What it shows is that varenicline reduces alcohol consumption, with an effect that looks larger than what we get with authorized treatments, but careful, that’s a comparison from a distance, not a head-to-head. Combining it with bupropion doesn’t do better than varenicline alone, bupropion alone is tested too late in the analysis plan for us to conclude anything from it, and the molecule’s only clear contribution is shortening the nausea. Where I practice, I’m not changing anything: neither drug has an approved indication for alcohol.”
Translated for practice: the article reads as a result about varenicline, not as a result about a combination. The trial’s title and what the trial actually establishes do not coincide, and it is this mismatch that makes the case worth attention. Keep the caution symmetrical too: neither the combination nor bupropion alone has been shown ineffective, they have not been shown effective, which is not the same thing.
What you can do with this
- Do not change alcohol use disorder practice on the basis of this trial. Authorization, marketing, and reimbursement are three separate decisions, and none of them is a verdict on efficacy for an indication a drug was never tested or licensed for. In France, status verified August 13, 2026: varenicline (CHAMPIX) holds a marketing authorization for smoking cessation in adults, has been marketed since its supply was confirmed by the ANSM on January 12, 2026, and is reimbursed at 65 percent; bupropion (ZYBAN SR 150 mg), authorized since August 3, 2001 as an aid to smoking cessation in nicotine-dependent patients, is marketed and not reimbursed. Neither drug is authorized for alcohol use disorder, where other agents are authorized in France and remain the reference framework. Prescribing outside a drug’s authorized indications is not automatically permitted: it depends on the absence of an appropriate authorized alternative, it engages the prescriber’s responsibility, and it must be justified, explained to the patient, and documented. Readers practicing outside France should check the equivalent rules in their own jurisdiction.
- Keep the reading lesson, which transposes to many combination trials: when a sequential testing hierarchy stops on a negative comparison, everything that follows remains descriptive, including p values below 0.05. And comparing two effect sizes each obtained against placebo never substitutes for the direct comparison between the two arms.
- In smoking cessation, for a patient bothered by prolonged nausea under varenicline, the idea of adding bupropion now has a randomized argument behind it. An argument to be handled with reservation: it comes from a secondary result, obtained in a different indication, in patients who were drinkers. It does not constitute a course of action.
- Pharmacological reminder, outside the scope of this trial: bupropion lowers the seizure threshold, which calls for particular vigilance in the alcohol context, especially during withdrawal, and its own contraindications must be checked before any combination. This is in fact why this trial’s protocol excluded any history of seizure. Varenicline has been the subject of a long debate over neuropsychiatric events, a debate that a thirteen-week phase 2 trial excluding significant psychiatric comorbidities could not settle.
- Watch for blood phosphatidylethanol as an endpoint. If it becomes established as a primary endpoint in alcohol trials, the field’s literature will become easier to read, since results will stop depending entirely on what patients report. This is probably the most durable contribution of this work.
Frequently asked questions
Does varenicline become a treatment for alcohol use disorder?
No. This is a phase 2 trial, conducted in a single country, over thirteen weeks, in a selected population. Authorization, marketing, and reimbursement are separate decisions from efficacy, and a phase 2 result, however clean, opens a development pathway, it does not amend an indication. In France, as of August 13, 2026, this molecule holds a marketing authorization for smoking cessation in adults, not for alcohol, an indication that has its own authorized treatments.
So the varenicline and bupropion combination has no value at all?
The precise formulation is more cautious. The trial did not demonstrate superiority of the combination over varenicline alone for reducing consumption, with p values of 0.97 and 0.76. This does not demonstrate that the two strategies are equivalent: the trial was not designed for that. The confidence interval of the difference is published and remains tight around zero on the biomarker, -0.123 to 0.119, which makes a substantial gain from the combination on this endpoint unlikely. What can be said is that nothing in this work justifies adding bupropion with the aim of improving the effect on consumption.
Is bupropion ineffective in alcohol use disorder?
That is not what the trial shows, and the same caution applies as to the previous question. The bupropion-alone arm is tested after the sequential hierarchy stopped: whatever its values, they cannot be interpreted as confirmatory. Its effect sizes on the two co-primary endpoints range from 0.14 to 0.28, while the trial was powered to detect an effect on the order of 0.43 to 0.44, and the appendix reports, for this arm, a few nominal p values below 0.05, uncorrected and uninterpretable. The honest conclusion is that this design could not settle the question of bupropion alone, not that it lacks an effect.
How good is blood phosphatidylethanol as an outcome measure?
It contributes something real: an objective measure of alcohol exposure over the preceding weeks, one that does not depend on what the patient reports, which is a notable advance in a field where endpoints are almost always self-reported. But it measures exposure, not disease. It says nothing about craving, functioning, somatic complications, or mortality. It is a surrogate marker of drinking behavior, objective but intermediate, and its translation into durable clinical benefit is not established by a thirteen-week trial.
Do the authors’ conflicts of interest invalidate the trial?
No, and it is important to stay factual. Five of eleven authors are co-owners of the company Sobrera Pharma AB, including the first author and the last author, who are its founders. These ties are disclosed. The sponsor is academic, the primary funding public, the analysis was conducted blinded on coded groups, an independent audit took place, and the authors state that the company neither produced nor supplied the study treatments and was not involved in either the statistical analysis or the data interpretation. An ownership conflict nonetheless remains the strongest degree of conflict of interest, and it weighs less on the numbers themselves than on how they are presented. And it is precisely the foregrounding of the combination, against what the direct comparison shows, that constitutes this article’s weak point.
Annotated bibliography
Source study. Söderpalm B, Lidö H, Franck J, Håkansson A, Lindqvist D, Heilig M, Guterstam J, Samuelson M, Askerup B, Wallmark-Nilsson C, de Bejczy A. Efficacy and safety of varenicline and bupropion, in combination and alone, for alcohol use disorder: a randomized, double-blind, placebo-controlled multicentre trial. The Lancet Regional Health Europe 2025; 54: 101310. DOI 10.1016/j.lanepe.2025.101310. PMID 40487775. Phase 2 randomized trial, four arms, 384 patients analyzed, thirteen weeks. Contribution: two monotherapy arms allowing the effect to be attributed, and a blood biomarker as co-primary endpoint. Limitations: likely functional unblinding through nausea, no imputation of missing data and uneven compliance across arms, testing hierarchy stopped at the central comparison, selected population, declared ownership conflict for five of eleven authors.
The trial protocol was published separately: de Bejczy A, Lidö H, Söderpalm B. A randomized, double-blind, placebo-controlled, multicentre trial on the efficacy of varenicline and bupropion in combination and alone for treatment of alcohol use disorder: Protocol for the COMB study. PLOS ONE 2024; 19: e0296118. DOI 10.1371/journal.pone.0296118. Registrations: EudraCT 2018-000048-24 and ClinicalTrials.gov NCT04167306.
What was consulted. The document initially provided by the publisher included only the notice page and the abstract. The full text, open access under a Creative Commons license, was consulted for this review: methods, randomization and blinding, sample size calculation, the sequential testing hierarchy, results, and the primary-results and safety tables, along with the PubMed record for the complete conflict-of-interest statement. The supplementary appendix was read in full, Tables S1 to S6 and the uncorrected secondary results, as was the trial’s published protocol. Not accessible: the participant flow diagram, the table of characteristics for the full population, the end of the discussion, and the end-of-article sections. The arm-by-arm study dropout figures and the limitations paragraph depend on these, and are therefore not reported here. Two internal discrepancies in the publication are flagged: the abstract and the results table give slightly different values for varenicline alone on the biomarker (d=0.30 and p=0.005 versus d=0.314 and p=0.004), and the abstract and the per-protocol table differ on several effect sizes. The values used here are those of the published abstract. French regulatory status was verified on August 13, 2026 against the public medicines database and the ANSM website. This article underwent an independent double reading.
Editorial collections
Tags
Verified on August 13, 2026 against the full text of the publication and its supplementary material where available. This analysis underwent an independent double reading. The English version was checked for conformity on September 23, 2026, against the figures of the French version and against the source. How we verify what we publish
This analysis is intended for healthcare professionals. It does not constitute a prescribing recommendation and does not replace individual clinical judgment.
Analysis from Psychiatry Evidence Base, evidence-based psychiatry, explained with rigor.