Paroxetine

Limits on what this shows

• Null/negative findings: No study found measuring mitochondrial parameters in normal (non-cancer, non-embryonic) adult human tissue at standard clinical paroxetine exposure.
• At-risk subgroups: The only population context directly modeled at a plausible exposure window is prenatal/developmental cardiac exposure (Shen et al., differentiating cardiomyocytes); not established for adult chronic dosing.

Regulatory status

Not addressed in FDA labeling; no mitochondrial-toxicity warning.

Population evidence

Steiner et al. 2015, in vitro, paroxetine interaction with mitochondrial proteins reported to mediate neuroprotection (mechanism-level, protective direction). Shen et al. 2025, in vitro, human pluripotent stem-cell-derived cardiomyocytes (2D and 3D cardiac organoid): paroxetine no-observed-adverse-effect level (NOAEL) 500 nM (differentiating 2D) to 1,000 nM (3D organoid) versus a clinical plasma concentration of about 55 nM (roughly 10-20x); "impaired mitochondrial basal and maximal respiration, and decreased spare respiration capacity" at NOAEL in both models; LC50 8.9 microM (H1-CMs) and 11.7 microM (hiPS-CMs). Rigotti et al. 2025, in vitro rat glial cells exposed to paroxetine plus quinolinic acid: mitochondrial dysfunction among the parameters measured, prevented by grape seed extract pretreatment (paroxetine concentration not available in the metadata accessed for this review). Wang et al. 2020, in vitro human lung cancer cell lines: paroxetine hydrochloride induced mitochondrial fragmentation and blocked autophagic flux (cancer-cell, high-concentration context; not representative of normal tissue at clinical dose).

Mechanism

Direct mitochondrial protein interaction, protective in the Steiner model; impaired oxidative phosphorylation machinery at high concentration in the Shen, Rigotti, and Wang models. Concentration- and cell-type-dependent, not a single consistent direction.

Counter-evidence held on record

Findings depend heavily on concentration and cell type; nothing here establishes mitochondrial harm at standard adult therapeutic dosing. This entry should not be read as a basis for CoQ10 or other mitochondrial-support supplementation, for which no dosing evidence exists in this drug's population. See Overlap map: the Shen and Rigotti sources are shared with the ATP production entry, and Rigotti is additionally shared with Oxidative stress.

Limits on what this shows

• Null/negative findings: None identified.
• At-risk subgroups: None established in humans; all direct evidence is animal or in vitro.

Regulatory status

Not addressed in FDA labeling.

Population evidence

Chung, Kim & Jin 2010, in vivo rodent model of Parkinson's disease: paroxetine prevented loss of nigrostriatal dopaminergic neurons "by inhibiting brain inflammation and oxidative stress" (direction confirmed at the title/indexed level; full quantitative data not independently retrieved for this review). Rigotti et al. 2025, in vitro rat glial cells: paroxetine plus quinolinic acid exposure increased reactive-oxygen-species/oxidative-stress markers, prevented by grape seed extract pretreatment. Erdemir et al. 2014, in vivo rat testicular tissue, measured oxidative stress parameters after paroxetine, sertraline, fluoxetine, and escitalopram administration; full text is blocked (SciELO returned HTTP 403 on both the HTML and PDF routes, no open-access mirror found via Unpaywall) and direction/magnitude could not be independently verified for this dossier. Listed for completeness in Sources with the access limitation flagged; not used to support this grade.

Mechanism

Not established at the systemic or human level; the two verified directions come from different tissues (dopaminergic neurons vs. glial cells) and different disease/toxicity contexts, not from a repeated finding in one system.

Counter-evidence held on record

Do not read this as either an antioxidant benefit or an oxidative-stress harm at the whole-body, therapeutic-dose level. The Chung 2010 protective finding is itself counter-evidence to any "paroxetine causes oxidative stress" claim. See Overlap map for the Rigotti source shared with Mitochondrial burden and ATP production.

Limits on what this shows

• Null/negative findings: None found.
• At-risk subgroups: None established.

Regulatory status

Not addressed in FDA labeling.

Population evidence

Gong et al. 2020, Neuroscience Letters, in vivo rat depression model. Title (confirmed via OpenAlex bibliographic record): "Joint replenishment of zinc and folic acid enhances the anti-depressive effect of paroxetine via increasing serum calcium and copper and decreasing serum arsenic." The article is paywalled with no open-access location returned by Unpaywall; only the title and bibliographic metadata could be verified for this dossier, disclosed here as an access limitation rather than a genuine absence of the study. A related study from an overlapping research group (Dou et al. 2018, Neuroscience Letters, PMID 30056106) used paroxetine as a comparator arm in a similar zinc/folic-acid rat design; its independent findings on paroxetine alone were likewise not retrievable.

Mechanism

Not established. The study design (zinc and folic acid added on top of paroxetine) cannot isolate any effect of paroxetine alone on serum copper. The general mechanism screen (Step 1/2) found no chelation, gastric-pH, or renal-wasting pathway by which paroxetine would plausibly alter copper status on its own.

Counter-evidence held on record

This is a single, access-limited, confounded study. It does not establish that paroxetine alone changes copper status, and no copper dosing recommendation follows from it.

Limits on what this shows

• Null/negative findings: None identified; no dedicated study exists to report a null result from.
• At-risk subgroups: None established.

Regulatory status

Not addressed in FDA labeling.

Population evidence

Lutokhina et al. 2022, Genes (Basel), human case report of mixed-etiology restrictive cardiomyopathy: the paroxetine-treated patient had elevated serum iron. Cardiomyopathy and heart failure are themselves well-documented causes of disturbed iron handling via hepcidin dysregulation, making attribution to paroxetine highly uncertain (n=1, no control, disease-confounded). Morse et al. 2004, J Mol Neurosci, in vitro screen of FDA-approved drugs for effects on amyloid precursor protein (APP) translation via the iron-responsive element in the APP 5' untranslated region: paroxetine was reported among compounds that "changed steady-state levels of transferrin" in this reporter-assay system, a translational-regulation readout, not a direct measure of iron absorption, storage, or classical iron status.

Mechanism

None established for classical iron depletion or excess. The mechanism screen found no absorption interference, no renal iron wasting, and no GI blood-loss signal specific to paroxetine in its label.

Counter-evidence held on record

The case report reflects one confounded patient, not a paroxetine effect; the in vitro study measured a translational-regulation endpoint, not iron status. Neither supports an iron-deficiency or iron-excess claim, and no iron dosing recommendation follows from either.

Limits on what this shows

• Null/negative findings: No significant beta-diversity change at either 7 or 14 days in the Dethloff study, despite the alpha-diversity signal; a partial null worth preserving alongside the positive finding.
• At-risk subgroups: None established in humans; all evidence is murine.

Regulatory status

Not addressed in FDA labeling.

Population evidence

Dethloff et al. 2020, Frontiers in Psychiatry, in vivo, male DBA/2J mice, n=17/group, paroxetine 5 mg/kg twice daily for 14 days: significantly reduced Faith's phylogenetic diversity (alpha diversity) at day 14 (P<0.04); no significant beta-diversity difference at 7 or 14 days; Firmicutes abundance correlated positively with behavioral improvement and negatively with paroxetine drug level, Bacteroidetes oppositely; 13 primary/secondary bile acids elevated at week 1 (3 significant) and 11 at week 2 (4 P<0.05, 2 P<0.01), including several sulfated bile acids notably elevated. Zhang et al. 2025, Basic & Clinical Pharmacology & Toxicology, in vivo mouse ulcerative colitis model, title: paroxetine "alleviates ulcerative colitis in mice via restoring intestinal microbiota homeostasis and metabolism" (restorative direction confirmed at the title/indexed level; full quantitative data not independently retrieved for this review).

Mechanism

Not fully characterized. Serotonin signaling has documented roles in gut motility, mucosal immune function, and bacterial gene expression, offering biological plausibility for both a diversity-reducing effect in a normal gut and a restorative effect in an inflamed one.

Counter-evidence held on record

All evidence is in mice; no human gut microbiome data exist for paroxetine. The direction of effect appears to depend on the baseline state of the gut, so this should not be generalized as a uniform depletion or harm claim.

Limits on what this shows

• Null/negative findings: None identified at therapeutic/clinical concentrations; both positive findings required concentrations well above typical steady-state plasma levels.
• At-risk subgroups: The only population context directly modeled is prenatal/developmental cardiac exposure (Shen et al.); not established for adult chronic dosing.

Regulatory status

Not addressed in FDA labeling.

Population evidence

Shen et al. 2025, Communications Biology, in vitro human pluripotent stem-cell-derived cardiomyocytes (2D differentiating and 3D cardiac organoid models): paroxetine NOAEL 500 nM (2D) to 1,000 nM (organoid) versus a clinical plasma concentration of about 55 nM; at NOAEL, "impaired mitochondrial basal and maximal respiration, and decreased spare respiration capacity and ATP production" in both models; LC50 8.9-11.7 microM. Rigotti et al. 2025, Molecular Neurobiology, in vitro rat glial cells exposed to paroxetine plus quinolinic acid: ATP levels among the parameters measured and altered, prevented by grape seed extract pretreatment (concentration not available in metadata accessed for this review).

Mechanism

Impaired mitochondrial oxidative phosphorylation at supratherapeutic concentration; mechanistically the same finding underlying the Mitochondrial burden entry (see Overlap map).

Counter-evidence held on record

Both supporting studies used concentrations well above typical clinical plasma levels. This does not establish an ATP-production deficit at standard adult dosing, and does not support CoQ10 or other ATP-targeted supplementation, for which no dosing evidence exists in this drug's population.

Limits on what this shows

• Why no grade: Both conditions for ungraded are met: no Tier B or Tier C evidence was found to support even a yellow grade, and no circulating consumer claim specific to paroxetine and oxygen delivery/hemoglobin/hematocrit was found on the sources checked. This is treated as a genuine evidence gap rather than a purely tooling-limited one, since the majority of search channels returned complete, clean negative results; the one server error is flagged so a later revision knows to re-check the OpenAlex channel specifically.

Channels tried

PubMed search-listing pages, four separately phrased queries: (1) paroxetine AND (hemoglobin OR hematocrit OR erythropoietin); (2) paroxetine AND erythrocyte membrane; (3) SSRI-class tissue oxygenation / mitochondrial respiration; (4) paroxetine AND (methemoglobinemia OR "oxygen saturation"). Also reviewed: FDA label adverse-reaction and clinical pharmacology sections (Step 1), consumer sources (Drugs.com, RxList) for any circulating oxygen/energy claim.

Outcome

Clean zero on all four completed PubMed searches: returned results were about glycated hemoglobin (HbA1c, a glucose marker, not oxygen-carrying hemoglobin), general hematocrit/oximetry physiology unrelated to paroxetine, or erythrocyte-membrane lipidomics used as a depression biomarker rather than an oxygen-delivery measure. One additional attempt via the OpenAlex works-search endpoint returned a server error (HTTP 503) rather than a clean negative; this single attempt is disclosed as tooling-limited, but the four PubMed searches that did complete returned genuine, complete result sets with no relevant hits.