The impact of levodopa on post-stroke depression: the ESTREL-depression-study

Clinical Question (PICO)

In adults with acute ischemic stroke or intracerebral hemorrhage and clinically meaningful hemiparesis requiring inpatient rehabilitation, does early levodopa/carbidopa, compared with matching placebo over a 39-day course, reduce depression-screen positivity three months after randomization?

The primary analysis selected participants without a documented history of depression who completed the questionnaire and took at least 80% of study medication. It therefore examines three-month symptom prevalence in an adherent, surviving subset—not treatment of established depression or definitively measured prevention of new-onset depression. The depression analysis was preplanned; the reported treatment-interaction analyses were exploratory, not established prespecified confirmatory subgroups.

Bottom Line

Levodopa did not demonstrate a reduction in post-stroke depressive symptoms: 56/209 participants (26.8%) versus 56/198 (28.3%) screened positive, OR 0.93 (95% CI 0.60–1.43). The authors conclude that levodopa is unlikely to be a suitable option for post-stroke depression during inpatient rehabilitation, but this result does not prove equivalence or answer whether it treats a diagnosed depressive episode. Primary results.

Design

Trial type: Preplanned exploratory depression analysis nested within ESTREL, a phase 3, multicenter, double-blind, parallel-group, placebo-controlled randomized trial.
N randomized in the parent trial: 610.
- Levodopa/carbidopa: 305.
- Placebo: 305.
N in the depression primary analysis: 407, selected after randomization.
- Levodopa/carbidopa: 209.
- Placebo: 198.
Randomization: Parent-trial allocation 1:1; no new randomization for this analysis. Stratification details are not specified in this depression report.
Setting: 13 acute stroke centers/units and 11 rehabilitation centers in Switzerland.
Enrollment: June 2019–May 2024.
Follow-up: Depression assessed at three months after randomization; a mean follow-up duration is not reported.
Analysis: Primary complete-case, on-treatment analysis using unadjusted binary logistic regression. Missing depression outcomes were not imputed. Sensitivity analyses included a 462-person available-outcome cohort regardless of adherence, inclusion of participants with prior depression, and post hoc adjustment/stratification for antidepressant use.
Primary outcome: PROMIS short-form depression-4a T-score ≥55, indicating at least mild depressive symptoms. The four items assess worthlessness, helplessness, depression, and hopelessness during the preceding week; this is a screening definition, not a psychiatric diagnostic interview.

Population

Inclusion Criteria

Adults with ischemic stroke or intracerebral hemorrhage within seven days before randomization.
Clinically meaningful hemiparesis; inpatient rehabilitation required and ability to participate in standardized therapy.
Randomization at least 24 hours after thrombolysis or thrombectomy, when these were performed.
Prestroke modified Rankin Scale (mRS) ≤3.
For this primary depression analysis: available three-month PROMIS data, ≥80% study-drug adherence, and no documented pre-existing depression.

Exclusion Criteria

Parent-trial exclusions included Parkinson disease, mandatory levodopa treatment, pre-existing hemiparesis, prestroke mRS >3, and severe aphasia preventing two-stage-command following.
Inability or unwillingness to comply with medication/study procedures; levodopa/carbidopa contraindications or hypersensitivity; pregnancy/breastfeeding and relevant contraception restrictions.
For this analysis: death or withdrawal before the assessment, incomplete/missing depression questionnaire, prior depression, or inadequate adherence.

The detailed parent eligibility and dose schedule above are verified against the trial registration. Prestroke mRS ≤3 should not be read as complete functional independence.

Baseline Characteristics (407-person analyzed cohort)

Median age 72 years (IQR 63–82); 60.4% male.
Ischemic stroke 86.0%; hemorrhagic stroke 14.0%.
Median NIHSS 7 (IQR 5–10); median poststroke mRS 4 (IQR 4–5).
Median Fugl-Meyer Motor Assessment (FMMA) 37 (IQR 17–58).
Aphasia 14.5%; severe aphasia was excluded from the parent trial.
Median time from stroke to randomization three days (IQR 2–5); first medication typically one day later.
Antidepressant use at baseline 8.8%, despite exclusion of documented prior depression; medication indications were not systematically available.

Interventions

Levodopa/carbidopa: 100/25 mg per administration, enterally by mouth or feeding tube, added to rehabilitation. The paper summarizes treatment as three times daily for 39 days; the registered schedule clarifies titration: once daily on days 1–3, twice daily on days 4–6, three times daily on days 7–34, twice daily on days 35–37, and once daily on days 38–39.
Placebo: Matching capsules with the same administration and titration/taper schedule, added to rehabilitation.
Background care: Rehabilitation in both groups. Antidepressants could be prescribed in clinical care; this was not a levodopa-versus-antidepressant trial.

Outcomes

Primary Outcome (PROMIS depression T-score ≥55 at three months):

Levodopa: 56/209 (26.8%); placebo: 56/198 (28.3%).
OR 0.93 (95% CI 0.60–1.43). A numerical P value is not reported for this comparison in the main results or Figure 2.
Calculated descriptive absolute difference: −1.49 percentage points, using the reported counts; this is not an adjusted treatment effect or a reported confidence interval.
The abstract rounds the levodopa proportion to 26%; the main figure gives 26.8%. Counts are the reference here.

Secondary Outcomes:

At least moderate depression (T-score ≥60): 29/209 (13.9%) versus 26/198 (13.1%), OR 1.07 (95% CI 0.60–1.89); numerical P value not reported.
Severe depression (T-score ≥70): 3/209 (1.4%) versus 0/198. No OR, confidence interval, or numerical P value is provided; these sparse counts do not establish harm.
Ordinal shift across no, mild, moderate, and severe depression: common OR 1.04 (95% CI 0.68–1.61); numerical P value not reported.
Sensitivity analysis including adherent participants with prior depression: 67/227 (29.5%) versus 62/210 (29.5%), OR 1.00 (95% CI 0.66–1.51).
The available-outcome analysis regardless of adherence and the post hoc antidepressant analyses were reported as consistent with the main result. Exact estimates are not supplied in the main-text narrative; no substitute estimates are invented here. Secondary and sensitivity results.

Exploratory Treatment Interactions:

Figure 5 reports no statistically significant interaction for age (P=0.523), sex (P=0.447), baseline FMMA (P=0.603), NIHSS (P=0.357), mRS (P=0.940), aphasia (P=0.719), or baseline antidepressant use (P=0.379). These are exploratory analyses; absence of an interaction is not proof that all subgroups respond identically. The figure below retains the reported subgroup estimates and confidence intervals.

Separate exploratory prognostic associations, not levodopa treatment effects, linked screen-positive depression to female sex (OR 2.11, 95% CI 1.36–3.28), baseline FMMA ≤37 (OR 1.84, 95% CI 1.18–2.88), and baseline NIHSS >7 (OR 1.86, 95% CI 1.20–2.91). These univariate associations do not establish causality or justify selecting patients for levodopa.

Adverse Events / Safety:

Participants with reported adverse events: 51/209 (24.4%) versus 34/198 (17.2%), P=0.10.
Participants with serious/severe adverse events as tabulated: 58/209 (27.8%) versus 63/198 (31.8%), P=0.43.
Depression recorded as an adverse event: 5/209 (2.4%) versus 5/198 (2.5%), P=1.00. This event-reporting category is not interchangeable with the much more frequent PROMIS screen-positive outcome.
These safety figures describe the selected analyzed cohort, not all randomized patients. The 28 deaths before the three-month visit are reported in aggregate, without arm-specific mortality estimates in this depression report. It cannot establish a mortality or intracranial-bleeding safety conclusion.

Figures

ESTREL-Depression participant selection flowchart — *Figure 1.* Study flow chart. Abbreviations: ITT = intention-to-treat, on-treatment analysis = analysis including the participants who took less than 80% of the study drug. The original caption mistakenly describes the on-treatment analysis as including participants taking less than 80% of study drug; the flowchart and Methods correctly show that these participants were excluded.

Three-month depression frequency and severity by randomized treatment — *Figure 2.* Forest plot of primary outcome and depression severity. The proportion of depression was measured using different cutoff T-scores of the PROMIS short-form depression-4a at three months, showing the number and proportion of participants for each levodopa and placebo group. Cutoff mild depression = T-score ≥ 55, cutoff moderate depression = T-score ≥ 60, cutoff severe depression = T-score ≥ 70.

Distribution of depression severity at three months — *Figure 3.* Distribution of the proportion of participants with different levels of depression severity in both treatment groups at 3 months. No depression = T-score < 55, mild depression = T-score ≥ 55 and < 60, moderate depression = T-score ≥ 60 and < 70, severe depression = T-score ≥ 70.

Sensitivity analysis including prior depression — *Figure 4.* Sensitivity analysis including all participants regardless of their history of depression. Exploring the proportion of depression using different cutoff T-scores of the PROMIS short-form depression-4a at three months with the number and proportion of participants each for the levodopa and placebo group. Cutoff mild depression = T-score ≥ 55, cutoff moderate depression = T-score ≥ 60, cutoff severe depression = T-score ≥ 70.

Exploratory treatment interaction forest plot — *Figure 5.* Interaction analysis of participant characteristics and the impact of levodopa on PSD. This interaction analysis shows how participant baseline characteristics are associated with the impact of levodopa on depression three months post-randomization. The median was chosen to convert each continuous numerical variable into a binary format. Abbreviations: FMMA = Fugl-Meyer Motor; LD = levodopa; mRS = modified rankin scale; NIHSS = National Institutes of Health Stroke Scale; PL = placebo.

Criticisms

Postrandomization selection: Only 407/610 randomized participants (66.7%, calculated) entered the primary analysis. The exclusions comprised 28 deaths, 11 withdrawals, 76 missing/incomplete questionnaires, 33 participants with previous depression, and 55 with inadequate adherence. Randomization does not automatically protect an adherent complete-case subset against selection bias.
Modified use of “ITT”: The report’s 462-person intention-to-treat sensitivity cohort removes the adherence restriction but still excludes deaths, withdrawals, missing outcomes, and prior depression. It is not an all-610 randomized analysis.
Prevalence is not incidence or treatment response: Depression was not assessed at baseline with the study scale. Removing documented historical depression does not establish absence of depression at randomization; neither new-onset prevention nor improvement of existing depression can be isolated.
Outcome measurement and external validity: Four affective screening questions are not a clinical diagnosis, and stroke-specific validation is limited. Exclusion of severe aphasia and missing questionnaires can underrepresent patients whose language or cognitive deficits make mood assessment difficult.
Timing and precision: Treatment lasted 39 days, but the questionnaire assessed the preceding week at three months. A transient on-treatment mood effect could be missed. The primary CI spans meaningful benefit and harm; this was not an equivalence trial.
Co-interventions: At three months, antidepressant use was 30.1% versus 22.2% (P=0.09). Sensitivity analyses were reassuring, but adjusting for a postrandomization clinical treatment cannot recreate an unconfounded levodopa-only biological experiment.
Exploration is not confirmation: Median-split subgroup variables, univariate prognostic models, and sparse severe-depression events invite overinterpretation. The critique here is this review’s methodological appraisal and the authors’ acknowledged limitations, not a claim that an external editorial was retrieved.

Funding

The parent ESTREL study was funded by the Swiss National Science Foundation (33IC30_179667). This depression report does not describe the funder’s operational role; it should not be assumed. Disclosures include institutional public/research grants and, for several investigators, outside-work industry fees or grants; the complete disclosure list is available in the paper. Registration: NCT03735901.

The paper

Authors. Mirjam I. Sauter et al.
Title. The impact of levodopa on post-stroke depression: the ESTREL-depression-study.
Journal. European Stroke Journal, 11(2):aakag001.
Year. 2026; published February 17, 2026. This October review is a look-back, not a claim of a newly published trial.
DOI. 10.1093/esj/aakag001
PMCID. PMC12911922
License. CC BY 4.0. Figures are reproduced unchanged with attribution to Sauter et al. and the original publication; no ownership is claimed.

Deep Dive — click to expand

What this is

This February 2026 paper tests the appealing idea that increasing dopamine during early stroke rehabilitation might also improve mood. Its randomized parent trial supplied a strong starting point, but the depression result comes from 407 adherent participants with available three-month questionnaires. The observed screen-positive rates were 26.8% and 28.3%, with an OR near one. The useful conclusion is that this regimen has no demonstrated mood benefit—not that dopamine is irrelevant to depression, or that every possible levodopa strategy has been disproved.

1. Shadow Audit

The shadow is the denominator: 203 of the 610 randomized participants did not reach the primary depression analysis. That is 33.3%, calculated—not an inconsequential amount of selection. A second shadow is the endpoint: participants were not assessed for baseline depression, so “post-stroke depression” here is a three-month screening state rather than a verified incident diagnosis.

There is also a clinically important measurement gap. Depression appeared in adverse-event reports in only 10/407 participants, while 112/407 screened positive. Those categories measure different things; the disparity does not prove undertreatment, but passive adverse-event recording clearly cannot substitute for systematic symptom assessment in this dataset. The negative drug result leaves the detection problem intact.

2. Inversion Engine

What would make the opposite conclusion—“routine early levodopa materially prevents depression”—credible? It would require a clinically worthwhile, reproducible reduction in prospectively defined depression, ideally with baseline assessment and analyses preserving allocation.

For illustration, reducing a 28% comparator risk to 18% would correspond to an OR of about 0.56, calculated. That is just beyond this study’s lower CI bound of 0.60. However, mapping the reported CI onto a fixed 28% comparator risk gives roughly 19%–36% treated risk, calculated: meaningful benefit and harm remain compatible with the data. This transformation is an illustrative scale translation, not a formal risk-difference CI. The estimate weakens a large universal benefit claim without proving a negligible effect. A targeted trial in patients with established depression or defined dopaminergic injury would ask a different question.

3. Second-Order Catalyst

The most plausible early adopters of this result are inpatient stroke-rehabilitation teams using levodopa mainly because it might improve both movement and motivation. The actionable change is to stop treating “dopamine supports motivation” as clinical evidence that this particular drug prevents depression.

Implementation does not require a new machine or biomarker. At medication review, document why levodopa is being prescribed and separate legitimate indications from speculative mood enhancement. Continue accessible mood screening and diagnostic follow-up; patients with aphasia or cognitive deficits require suitable assessment rather than being treated as screen-negative. This is an editorial implementation inference, not a trial-tested care bundle.

4. Asymmetric Leverage

The asymmetric payoff is de-implementation of an unproved add-on while preserving attention to a common, patient-important problem. Even among selected adherent survivors, about one quarter screened positive at three months. A protocol that removes an automatic medication order but retains active mood assessment could reduce treatment burden without abandoning care.

The trial did not measure cost savings, staff time, or the benefit of a screening program, so none should be claimed as demonstrated outcomes. The leverage is conceptual: direct limited rehabilitation attention toward an identified need instead of assigning a mechanistic promise to a drug with no demonstrated benefit on this endpoint.

5. Paradigm Destroyer

The reflex this paper challenges is moving directly from a neurotransmitter association to a routine prescription. A low dopamine level can be a marker or consequence of injury, not necessarily a modifiable cause that responds to precursor replacement.

Two-sentence protocol update: Do not initiate levodopa/carbidopa solely to prevent post-stroke depression on the strength of a dopamine/motivation rationale; retain other established indications and individual specialist decisions. Screen for depressive symptoms with methods suited to communication ability, confirm suspected depression clinically, and use a separate evidence-based treatment pathway rather than assuming rehabilitation augmentation treats mood.

This is a bounded appraisal of this regimen, not a comprehensive guideline update or advice to abruptly stop existing treatment.

MVP — Minimum Viable Proof

The minimum useful statement is numerical and narrow: among 407 selected participants, three-month depression-screen positivity was 56/209 versus 56/198, OR 0.93 (95% CI 0.60–1.43). That provides no affirmative evidence for making a 39-day early levodopa course a routine depression-prevention order. It does not establish equivalence, efficacy in diagnosed depression, or safety in all randomized patients.

Best Combination

Read this alongside the parent ESTREL motor-recovery result, which this paper reports as showing no benefit from levodopa added to standardized rehabilitation. Their alignment makes a dual motor-and-mood rationale less persuasive, although the depression analysis is less protected against selection and was not the parent trial’s primary outcome. The parent trial is Engelter et al., JAMA 2025; this contextual statement follows its description in the reviewed paper rather than a new quantitative reanalysis.

The authors also discuss earlier small trials of methylphenidate or combined methylphenidate/levodopa. Those heterogeneous interventions should not be pooled conceptually as proof of a levodopa class effect. A sensible synthesis separates motor recovery, depressive-symptom prevention, and treatment of diagnosed depression—and demands evidence for each rather than allowing a positive mechanism in one domain to migrate into another. No claim about current professional guideline endorsement is made here.

Overvalue Warning

Do not upgrade “no demonstrated effect” to “no possible effect.” The CI is broad; the endpoint was measured after the treatment course, and the primary cohort was selected. A brief on-treatment response, different regimen, or biologically defined subgroup remains untested.
Do not treat a positive screen as a psychiatric diagnosis, or this prevention-like analysis as a depression-treatment trial. Baseline symptoms were not measured, prior depression was excluded from the primary analysis, and severe communication deficits were underrepresented. The result neither licenses neglect of mood symptoms nor rules out other depression treatments.

Educational use notice. The Time Is Brain Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.