Paroxetine: the legal and safety record

The timeline that is the story

1. 2001. Study 329 — a GSK-funded trial of paroxetine in adolescents — is published, reporting the drug as generally well tolerated and effective [3]. Paroxetine was never FDA-approved for anyone under 18.
2. July 2012. GSK pleads guilty and pays $3 billion — then the largest healthcare-fraud settlement in US history. The criminal information on the Paxil count cites the unlawful promotion of the drug to patients under 18, including “preparing, publishing and distributing a misleading medical journal article” [1][2]. The plea is about conduct; it is not a finding that the drug harmed any particular patient.
3. September 2015. An independent team reanalyzes the original Study 329 data under the RIAT initiative and publishes the result in The BMJ: in that trial, paroxetine was not significantly better than placebo for adolescents on the prespecified primary outcomes, and harms were understated [4]. Same data, opposite conclusion.
4. September 2025. Twenty-four years after publication — and thirteen after the guilty plea — the journal attaches an expression of concern to Study 329, its first formal mark [5], three weeks after an amended complaint asking a Washington, DC court to order the paper retracted was filed against the journal’s owner and its publisher [8]. That suit is an allegation, not a finding, and we have not been able to read its current status. The paper has never been retracted and remains citable today.

That last fact is the reason this page exists: the correction machinery of the scientific literature moved slower than the Department of Justice, by more than a decade. When we say every study in the evidence library gets its funding read literally and its retraction status checked, Study 329 is why the check cannot stop at “not retracted.”

The birth-defect litigation, stated precisely

GSK settled most Paxil birth-defect claims without saying what it paid. Its annual filings with the SEC say that by February 2012 it had agreed to settle “the substantial majority” of US claims over Paxil use in pregnancy, that other claims were dismissed without payment, and that the first trial, Kilker in Philadelphia in October 2009, ended in a $2.5 million verdict for the plaintiff with no punitive damages [7]. (An earlier version of this page repeated a 2010 Bloomberg report that GSK paid more than $1 billion across more than 800 cases. No GSK filing states a total, so that figure is gone.) No court ever adjudicated general causation against GSK on birth defects. The contrast case is instructive: the equivalent sertraline litigation was decided for the manufacturer and affirmed on appeal — that record is here. Settled and decided-against are different legal outcomes, and most coverage of “SSRI birth defect lawsuits” blurs them into one.

What is not in the record

• No Class I recall (within openFDA's ~2012-onward coverage window). Twenty-three recall records exist for paroxetine — twenty-two Class II, one Class III, four ongoing — all for manufacturing and quality problems [6]. FDA’s most serious category has never been applied.
• No adjudicated causation finding on birth defects — see above; GSK never disclosed the settlement amounts and the record stops there [7].
• No retraction of Study 329. An expression of concern is a warning label, not a withdrawal; the paper stands in the literature [5]. We state this as a fact about the literature’s correction machinery, in both directions.

What this means if you take paroxetine

The record above concerns adolescent promotion and corporate conduct from two decades ago — it is not a verdict on whether paroxetine is right for an adult taking it now, and it changes nothing about your prescription by itself. Two practical notes do follow from it. First, paroxetine has one of the strongest withdrawal profiles among SSRIs, so stopping abruptly in reaction to a page like this is the single worst available move — any stop belongs in a planned taper with your prescriber. Second, the fair question this history hands you is about evidence, not danger: “how much of what we know about this drug comes from manufacturer-run trials, and what independent evidence do we have?” Our paroxetine page carries the funding line on every study for exactly that conversation.

Common questions

What was GSK’s $3 billion settlement actually for?

In July 2012, GlaxoSmithKline pleaded guilty to criminal charges and paid $3 billion — at the time the largest healthcare-fraud settlement in US history. The criminal plea covered three drugs: promoting Paxil for patients under 18 when it was never approved for them, promoting Wellbutrin for unapproved uses, and failing to report safety data on Avandia. The Paxil count specifically cited the preparation and distribution of a misleading journal article about adolescent use.

What is Study 329?

A GSK-funded trial of paroxetine in adolescents, published in 2001, that reported the drug was generally well tolerated and effective. The DOJ’s 2012 criminal information described the resulting journal article as misleading. In 2015, an independent reanalysis of the same trial data under the RIAT initiative (Restoring Invisible and Abandoned Trials, which re-publishes trials from their raw data), published in The BMJ, concluded paroxetine was neither effective nor safe for adolescents in that trial.

Was Study 329 ever retracted?

No. Despite the guilty plea and the published reanalysis, the paper stood unmarked for twenty-four years. In September 2025 the journal attached an expression of concern — a flag short of retraction — three weeks after an amended complaint in a Washington, DC consumer-protection suit asked a court to order the paper retracted. That suit is an allegation, not a finding. The paper remains citable today.

Did courts find that Paxil causes birth defects?

No court adjudicated general causation against GSK. GSK told the SEC that by February 2012 it had agreed to settle the substantial majority of US claims over Paxil use in pregnancy, with other claims dismissed without payment; it has not published what it paid. The first trial, Kilker in Philadelphia in October 2009, ended in a $2.5 million verdict for the plaintiff, with no punitive damages. Contrast sertraline, where the birth-defect MDL (multidistrict litigation — see /glossary/mdl) was decided for the manufacturer and affirmed on appeal; settled and decided-against are different outcomes, and paroxetine’s record is the former.

Has paroxetine had serious recalls?

No Class I recall — FDA’s most serious category — appears in the openFDA enforcement database (openFDA enforcement data is reliable only from roughly 2012 onward, so this is a statement about the covered window, not the drug’s whole history). Twenty-three records exist: twenty-two Class II and one Class III, four still ongoing, for manufacturing and quality problems rather than anything about the molecule itself.

Does any of this mean I should stop taking paroxetine?

No. The record above is about corporate conduct and adolescent prescribing two decades ago, not about whether the drug is right for an adult taking it today — and paroxetine has one of the strongest withdrawal profiles among SSRIs, which makes stopping abruptly the single worst response to reading this page. Bring questions to your prescriber, including a planned taper if you ever do stop.

Sources

Last verified 2026-09-29. Corrections change this date.