Fact-checked by Grok 2 weeks ago

Study 329

Study 329 was a multicenter, double-blind, randomized, -controlled sponsored by SmithKline Beecham (later GlaxoSmithKline) from 1994 to 1998, evaluating the efficacy and safety of versus and in 275 adolescents aged 12 to 18 with unipolar major depression. The trial featured an 8-week acute phase assessing response via the Kiddie-Schedule for Affective Disorders and Schizophrenia (K-SADS), followed by a 6-month continuation phase. Published in 2001 by Keller et al. in the Journal of the American Academy of , the study reported as generally well tolerated and effective for reducing depressive symptoms compared to . However, internal analyses and full clinical study reports revealed paroxetine failed all protocol-specified efficacy outcomes, showing no statistically significant benefit over . Reanalysis in 2015 by LeNoury et al., accessing through legal channels, confirmed the absence of efficacy while identifying increased harms, including and behavior (8 events on paroxetine versus 1 on ) and other serious adverse events. These discrepancies, including selective reporting and ghostwriting allegations, contributed to U.S. of findings of fraudulent promotion by GlaxoSmithKline, resulting in a $3 billion settlement in for misleading safety claims on paroxetine in . The trial's fallout prompted the FDA to issue a warning in 2004 on all antidepressants for increased suicidality risk in children and adolescents, informed in part by Study 329's unreported events. It exemplifies challenges in pharmaceutical trial , with advocates citing it to push for mandatory disclosure beyond summaries. Recent scrutiny, including a 2025 expression of concern on the original publication, underscores ongoing debates over and retraction standards.

Clinical Trial Design and Execution

Objectives and Protocol

Study 329, sponsored by SmithKline Beecham (later GlaxoSmithKline), aimed to evaluate the efficacy and safety of , an , relative to and , a , for treating unipolar in adolescents. The trial specified two coprimary efficacy endpoints for the 8-week acute phase: the proportion of treatment responders, defined as participants achieving a total score of ≤8 on the 17-item (HAM-D) or a reduction of ≥50% from baseline HAM-D score at endpoint; and the mean change from baseline to endpoint in total HAM-D score. Secondary objectives encompassed additional measures of depressive symptoms, global improvement, and suicidality risk, with the delineating five secondary efficacy endpoints alongside safety monitoring for adverse events, including emergent suicidality. The protocol outlined a multicenter, randomized, double-blind, -controlled, parallel-group conducted across 12 sites in the United States from 1994 to 1998. Eligible participants were outpatients aged 12-18 years meeting DSM-IV criteria for , with a baseline HAM-D score ≥20 or Clinical Global Impression-Severity score ≥4, and no significant comorbidities or recent psychotropic use. Following a 7- to 14-day screening and washout period, 189 adolescents were randomized in a 2:2:2 ratio to receive flexible dosing of (10-60 mg/day, targeting 20-40 mg), (titration to 100-200 mg/day), or matching over 8 weeks, with weekly clinic visits for assessments using scales such as the HAM-D, Montgomery-Åsberg Depression Rating Scale, and Children's Depression Rating Scale-Revised. Responders could enter an optional 6-month open-label continuation phase for long-term safety and relapse prevention data, though this was not powered for efficacy comparisons.

Methods and Participant Characteristics

Study 329 was a multicenter, double-blind, randomized, placebo-controlled, parallel-group evaluating the efficacy and safety of compared to and in adolescents with . The acute phase lasted eight weeks and was conducted at 12 academic centers in (10 in the United States and 2 in ) from April 20, 1994, to February 15, 1998. Participants underwent a screening period of 7 to 14 days, followed by in a 1:1:1 ratio using a computer-generated list balanced in blocks of 6 or 8. Interventions included flexible dosing of (20-40 mg/day, titrated over four weeks with possible increases for non-responders), (starting at 50 mg/day up to 300 mg/day), or matching capsules, alongside weekly sessions. Assessments occurred weekly using standardized instruments, including the (HAM-D), (CGI) scale, and Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS-L) for diagnostic confirmation. Eligibility required outpatients aged 12-18 years meeting DSM-IV criteria for non-psychotic of at least eight weeks' duration, a HAM-D total score of 12 or higher, a (CGAS) score below 60, and an estimated IQ of 80 or above via the . Exclusion criteria encompassed serious suicidality with intent or plan, , , , or dependence, certain comorbid conditions (e.g., eating disorders, obsessive-compulsive disorder, , within the past year), recent use, or medical contraindications to the study drugs. Of 425 screened adolescents, 275 were randomized: 93 to , 95 to , and 87 to . Baseline characteristics were similar across groups, with mean ages ranging from 14.8 to 15.1 years (standard deviation ±1.6 years), approximately 60% , and 80-87% participants. Mean HAM-D scores were 18.1-19.0, reflecting moderate severity, and mean episode duration was about 14 months, with 41-50% having at least one comorbid psychiatric (most commonly anxiety disorders in 19-28%). Many participants had chronic exceeding one year.

Conduct and Data Collection

Study 329 was a multicenter, double-blind, randomized, placebo-controlled, parallel-group sponsored by SmithKline Beecham (later GlaxoSmithKline), conducted across 12 North American sites—10 in the United States and 2 in —from April 20, 1994, to February 15, 1998. Participants were 275 adolescents aged 12-18 years diagnosed with unipolar per DSM-III-R criteria, requiring a Hamilton Depression Rating Scale (HAM-D) total score of at least 12 and symptoms persisting for no less than 8 weeks. Following a 7- to 14-day screening period to confirm eligibility and exclude conditions such as suicidality, , or significant comorbidities, eligible participants were randomized in a 2:2:1 ratio to receive (initially 10-20 mg/day, titrated to 20-40 mg/day), (initially 25-50 mg/day, titrated to 200-300 mg/day), or matching over an 8-week acute phase, with an optional 6-month continuation phase for responders. The trial adhered to the protocol, which underwent amendments in 1993, 1994, and 1996, and was executed by investigators from university-affiliated and hospital psychiatry departments. Data collection occurred through standardized forms (CRFs) completed at each site, encompassing approximately 77,000 pages across all participants, with assessments scheduled at and weekly for the first four weeks, then at weeks 6 and 8. were gathered using validated instruments, including the 17-item HAM-D for total score change and responder status (defined as ≥50% reduction from or endpoint score ≤8), supplemented by depression-specific items from the Kiddie Schedule for Affective Disorders and (K-SADS-L), (CGI) scales, and secondary measures such as the Scholastic Performance Profile (SPP) and Adolescent Family Conflict (AFC). Safety monitoring involved systematic recording of adverse events elicited through open-ended investigator queries, alongside , tests, electrocardiograms (EKGs), and plasma/ drug levels at weeks 4 and 8; treatment compliance was verified via capsule counts. All were compiled for intent-to-treat analysis using last observation carried forward (LOCF) imputation, with categorical outcomes analyzed via and continuous measures via analysis of variance (ANOVA). The protocol specified a taper phase for all completers and a 30-day post-treatment follow-up for safety monitoring, though execution focused primarily on the acute phase data for initial reporting. No significant deviations in trial conduct were reported in contemporaneous documents, though later independent access to raw data highlighted inconsistencies in adverse event transcription from CRFs to summary reports.

Original Results and Internal Analyses

Efficacy Outcomes

The protocol for Study 329 specified two co-primary efficacy endpoints: the change from baseline to week 8 in the total Hamilton Depression Rating Scale (HAM-D) score, and the percentage of responders defined as a Clinical Global Impression-Improvement (CGI-I) score of 1 ("very much improved") or 2 ("much improved"). Secondary efficacy outcomes included changes in depression symptoms measured by the Kiddie-Schedule for Affective Disorders and (K-SADS), CGI-Severity (CGI-S), and responder rates based on a ≥20% reduction in HAM-D total score or K-SADS depressed mood item. In the original analysis reported by GlaxoSmithKline (GSK), (20-40 mg/day) did not demonstrate superiority over on the HAM-D total score change (-9.0 for vs. -10.4 for ; not statistically significant). However, the publication claimed efficacy based on the CGI-I responder rate, reporting 66% for versus 47% for (p=0.002, unadjusted). showed no superiority on either co-primary endpoint compared to . GSK internal documents from 1998, revealed through litigation, concluded that the study "failed to demonstrate a statistically significant difference from on the primary efficacy measures," labeling it a negative for efficacy on all eight protocol-specified outcomes. Reanalysis of the full under the Restoring Invisible and Abandoned Trials (RIAT) initiative confirmed no for or over across all nine pre-specified outcomes, including both co-primaries and key secondaries like K-SADS and CGI-S changes. The apparent CGI-I finding in the original report relied on selective reporting of a single secondary measure without adjustment for multiplicity or deviations, such as post-hoc handling of dropouts; when analyzed per with all data, no significant benefits emerged (e.g., CGI-I 1.4, 95% CI 0.8-2.5, p=0.23). These discrepancies highlight how outcome selection influenced the published claim of , despite internal recognition of failure.

Safety and Adverse Events

In Study 329's acute phase, treatment was linked to elevated rates of serious adverse events (SAEs) relative to , with 11 of 93 participants (11.8%) on experiencing SAEs—9 leading to discontinuation—compared to 2 of 87 (2.3%) on and 5 of 95 (5.3%) on . The original 2001 publication reported as generally well tolerated, noting adverse event-related discontinuation in 9.7% of participants versus 6.9% on , without emphasizing the disparity in SAEs or classifying certain events (e.g., severe psychiatric issues, 32 under ) as clinically significant harms. Suicidality emerged as a key concern in the reanalysis of the , revealing suicidal or self-injurious behaviors in 11 participants (11.8%) versus 2 on (2.3%)—a rate over fivefold higher—while showed 4 cases (4.2%); this contrasted with the original report's lower counts of 5, 1, and 3 events across groups, respectively, which aggregated or omitted some ideations and gestures from raw data. Overall psychiatric adverse events totaled 103 under but only 24 with , including heightened incidences of , , and . The reanalysis concluded clinically meaningful harm elevations with , including these suicidality risks and other SAEs not deemed superior to in internal GSK assessments reflected in the full , though the original downplayed such patterns by selective outcome selection and coding. Total adverse events were also higher (481 MedDRA-coded under versus 330 on ), with underreporting in the appendices by up to 14%.

GSK Internal Interpretations

GSK's internal clinical study synopsis for Study 329, prepared following database lock in 1998, concluded that did not achieve on either of the two pre-specified primary endpoints: the mean change from to week 8 in the total (HAM-D) score (-10.7 for versus -8.9 for , p=0.113) and the proportion of treatment responders based on the (CGI) Improvement score (67% for versus 55% for , p=0.112). , the active comparator, also failed to separate from on these measures. Despite these results, GSK highlighted statistically significant improvements on select secondary endpoints, including the depressed mood item of the HAM-D (p=0.003) and the K-SADS-L item (p=0.049), interpreting the overall findings as demonstrating a "modest benefit" of over for adolescent major . On safety, the internal synopsis reported 11 serious adverse events (SAEs) in the group compared to 2 in the group and 5 in the group, with SAEs in the arm predominantly involving psychiatric issues such as , , worsening , and intentional overdose. GSK noted that 's safety profile was "generally similar" to that observed in adults, though additional adolescent-specific adverse events like tooth disorder and hostility were identified; common treatment-emergent adverse events for included (24%), (17%), and (15%). Internally, these findings contributed to GSK's acknowledgment that Study 329 and related trials provided insufficient evidence to support FDA approval of for pediatric . An internal SmithKline Beecham (predecessor to GSK) memorandum from 1998 emphasized the commercial risks of the negative primary results, directing efforts to "effectively manage the dissemination of these data in order to minimise any potential negative commercial impact." This approach involved prioritizing secondary analyses for publication while de-emphasizing the failed primaries and elevated SAE rates, reflecting GSK's strategic framing of the study as supportive of paroxetine's utility despite the absence of robust efficacy signals.

Publication Process

Authorship and Ghostwriting

The manuscript reporting Study 329 was drafted primarily by Sally K. Laden, a medical writer from , a firm contracted by SmithKline Beecham (SKB, now GlaxoSmithKline or GSK) to prepare the publication. STI received $17,250 for producing up to six drafts based on SKB's internal synopsis, with Laden acknowledged in the final article only for "editorial assistance"—a phrase often used to obscure substantive ghostwriting contributions in pharmaceutical-sponsored research. Listed as lead author was Martin B. Keller of , accompanied by 21 co-authors, including principal investigators like Neal D. Ryan and Michael Strober, many of whom contributed primarily through site-level or coordination rather than development. Internal SKB documents, disclosed during subsequent litigation, reveal that Keller originated the as a key opinion leader but did not produce the initial draft, despite his public assertions to the contrary; Laden confirmed drafting it independently under SKB guidance. SKB's central medical affairs team exerted significant control, with employees such as Donna McCafferty and Murray Oakes directing 11 iterations of the manuscript from 1998 to 2000 to emphasize efficacy signals and downplay adverse events, aiming to offset potential commercial harm from the negative results. , as data owner, required all content to undergo medico-legal before release to academic authors for journal submission, which proceeded first to (rejected) and then to the Journal of the American Academy of (JAACAP), where Laden handled responses leading to acceptance and publication on July 1, 2001.

JAACAP Article Content

The Journal of the American Academy of Child and Adolescent Psychiatry (JAACAP) article, published in July 2001 and titled "Efficacy of Paroxetine in the Treatment of Adolescent Major Depression: A Randomized, Controlled Trial," presented Study 329 as a multicenter, double-blind, randomized, placebo- and active-controlled parallel-group trial evaluating paroxetine hydrochloride (Paxil) against placebo and imipramine in adolescents with unipolar major depressive disorder. The trial enrolled 275 outpatients aged 12 to 18 years meeting DSM-IV criteria, with randomization to paroxetine (n=189, mean endpoint dose 30.3 mg/day), imipramine (n=92, mean endpoint dose 236.6 mg/day), or placebo (n=94).60309-9/abstract) Treatment lasted 8 weeks, with assessments using the Hamilton Rating Scale for Depression (HAM-D, 17-item), Kiddie Schedule for Affective Disorders and Schizophrenia-Lifetime Version (K-SADS-L) depression section, and Clinical Global Impression (CGI) scales. The article reported efficacy primarily through secondary endpoints, stating that paroxetine yielded statistically significant improvements over placebo in achieving a HAM-D total score of ≤8 at endpoint (17% vs. 6%, p=0.02), the HAM-D depressed mood item score (p<0.001), the K-SADS-L depressed mood item score (p=0.006), and CGI-improvement responder status (75% vs. 61%, p=0.002). Imipramine showed no significant differences from placebo on these or primary measures like HAM-D total change or responder rates (≥50% reduction), though a post-hoc combined active treatment analysis suggested overall superiority to placebo (p=0.02 for HAM-D responders).60309-9/abstract) The authors described paroxetine's effects as consistent with adult SSRI trials and positioned the results as extending evidence for selective serotonin reuptake inhibitors (SSRIs) to adolescents. Safety data in the article emphasized general tolerability comparable to adult populations, with the most frequent adverse events for paroxetine including asthenia (14%), insomnia (13%), and headache (12%), versus placebo rates of 8%, 9%, and 16%, respectively.60309-9/abstract) Serious adverse events occurred in 5 paroxetine patients (2.6%), 5 imipramine patients (5.4%), and 2 placebo patients (2.1%), including one hospitalization for dehydration on paroxetine; investigators attributed none to the study drug. Suicidal ideation or gestures were noted in 6 paroxetine, 2 imipramine, and 1 placebo patient, with no completed suicides across groups.60309-9/abstract) Cardiovascular parameters showed no clinically meaningful changes, though paroxetine patients experienced mean decreases in supine heart rate (-4.8 bpm) and diastolic blood pressure (-2.1 mm Hg); a small mean weight loss of 1.2 kg was observed across all arms. The conclusions asserted that "paroxetine is generally well tolerated and effective for major depression in adolescents," recommending it as a first-line treatment option while calling for further long-term studies on maintenance and relapse prevention. The article highlighted the trial's flexibility in dosing (paroxetine 10-60 mg/day after initial titration) and exclusion of patients with suicidality at baseline, framing these as strengths for real-world applicability.60309-9/abstract)

Discrepancies with Raw Data

The original 2001 publication in the Journal of the American Academy of Child & Adolescent Psychiatry (JAACAP) reported paroxetine as generally well tolerated and effective for adolescent major depression, citing improvements on several secondary measures such as the Clinical Global Impression-Improvement (CGI-I) responder status and certain Hamilton Depression Rating Scale (HAM-D) subscales. However, reanalysis of the full clinical study report (CSR) and raw data revealed that the protocol-specified primary efficacy endpoint—change in total HAM-D score from baseline to endpoint—showed no statistically significant difference between paroxetine and placebo (mean decrease of 10.7 points for paroxetine versus 9.1 points for placebo; P=0.20), falling short of the prespecified clinical significance threshold of a 4-point difference. The publication emphasized four non-protocol-specified depression scales as evidence of efficacy, representing selective outcome reporting that deviated from the original protocol's nine prespecified efficacy measures, none of which demonstrated superiority of paroxetine or imipramine over placebo. On safety, the JAACAP article described paroxetine as having a profile comparable to placebo, listing 265 adverse events without highlighting increased risks, and minimized psychiatric adverse events by categorizing suicidal ideation or gestures under less severe terms. In contrast, the raw data documented 481 adverse events for paroxetine, including 11 serious adverse events (versus 2 for placebo), with five suicidal acts or ideation events on paroxetine compared to one on placebo; these included behaviors such as overdoses and preparatory acts not emphasized in the publication. Reanalysis confirmed higher withdrawal rates due to adverse effects (15% for paroxetine versus 6.9% for placebo) and an overall increase in harms, including aggression and akathisia, which the original article omitted or downplayed through post-hoc classifications that excluded protocol-defined severe events. Statistical discrepancies further underscored misrepresentation: the publication relied on pairwise comparisons without establishing overall omnibus significance (e.g., ANOVA F-test P>0.05 across groups), selectively reporting favorable p-values from secondary analyses while ignoring nonsignificant primary results and multiplicity adjustments. For instance, the claimed was a post-hoc not prioritized in the , and hazard ratios for remission favored in unadjusted models. The reanalysis, adhering to the CSR and without such switches, found no of and confirmed elevated signals, attributing the original deviations to outcome switching and selective emphasis inconsistent with the raw dataset.

Marketing and Promotional Use

Off-Label Promotion Strategies

GlaxoSmithKline (GSK) promoted (Paxil) for in treating in children and adolescents by distributing reprints of the 2001 Journal of the American Academy of Child & Adolescent Psychiatry (JAACAP) article on Study 329 to physicians through its sales representatives. These materials selectively highlighted claims of and tolerability from secondary outcomes, despite the study's failure to meet its primary endpoints for and internal analyses showing no benefit over . Sales detailing emphasized paroxetine's supposed advantages for adolescent patients, positioning it as a viable option in a market where the drug lacked FDA approval for individuals under 18. To further incentivize off-label prescribing, GSK provided free Paxil samples to physicians who primarily treated pediatric patients, a practice documented in internal records as part of broader efforts to penetrate practices. This distribution occurred alongside verbal promotions during sales visits, where representatives cited Study 329's published results to counter concerns about unapproved uses, even as GSK's own data indicated increased risks of suicidality and behavioral adverse events in youth. Such tactics contributed to a reported surge in off-label prescriptions, with U.S. scripts for adolescents rising significantly following the article's release in July 2001. GSK's promotional activities extended to internal sales strategies that framed Study 329 as supportive evidence for expanding paroxetine's indications, despite regulatory prohibitions on off-label under the Food, Drug, and Cosmetic Act. A 2001 internal sales acknowledged the off-label potential, noting plans to leverage the JAACAP for pediatric amid awareness that primary trial data did not justify approval submissions to the FDA. These efforts persisted until 2003, when Eliot Spitzer's investigation revealed suppressed suicidality data, prompting GSK to issue a "" letter on June 10, 2003, acknowledging heightened risks in youth without conceding inefficacy. The strategies culminated in GSK's 2012 guilty plea to misdemeanor misbranding charges, including off-label of Paxil, resulting in a $3 billion settlement—the largest in U.S. healthcare fraud history at the time.

Internal Sales Documents

GSK's internal sales training materials for Paxil misrepresented Study 329 as evidence of the drug's efficacy and safety for adolescent depression, despite internal analyses concluding the trial failed its primary and secondary endpoints. In September 1999, sales representatives received training from Dr. Karen Wagner, a paid consultant, who presented selective data from Study 329 to emphasize positive outcomes while omitting the lack of statistical significance on key measures like Hamilton Depression Rating Scale improvements and response rates. These sessions targeted a 150-person neuroscience specialty sales force, instructing reps to promote off-label use to physicians treating patients under 18. Promotional aids distributed to approximately 2,000 sales representatives included an August 16, 2001, cover memo accompanying the article on Study 329, which highlighted "remarkable efficacy" claims but concealed negative results and the FDA's non-approval for pediatric use. Sales call notes, accessible to supervisors, documented representatives citing the article and Wagner's lectures to advocate Paxil for adolescent depression, often without disclosing increased risks such as suicidality. GSK's Paxil Forum events from 2000 to 2002, held at resorts and attended by child psychiatrists, featured slides and presentations by Wagner that selectively portrayed Study 329 data to support expansion into pediatric markets, aligning with a 2000 internal consultant report recommending such positioning despite known trial shortcomings. Training manuals and seminars, such as the December 5, 2000, TSR Representatives Training, further encouraged off-label promotion by framing Study 329 alongside Studies 377 and 701 as supportive of Paxil's benefits, burying findings of no and up to threefold increased risk. FaxBack materials and comparative documents, like those in Wellbutrin Semester II training (2001), lacked fair balance and promoted Paxil as "generally well-tolerated and effective" for children, contributing to fraudulent claims under federal health programs. An August 2002 strategic brand plan internally identified pediatric depression as a sales opportunity, guiding these materials toward maximizing prescriptions despite regulatory constraints.

Regulatory Context of Promotion

In the United States, the (FDA) regulates pharmaceutical under the Federal Food, , and Cosmetic Act (FD&C Act), specifically 21 U.S.C. § 352, which deems a drug misbranded if its labeling or includes unapproved uses, false or misleading claims about safety or efficacy, or failure to reveal material facts. is confined to FDA-approved indications listed in the drug's labeling, with off-label —disseminating information to encourage use beyond approved populations, dosages, or conditions—prohibited as it risks without established benefit-risk data. Violations can trigger FDA letters, product seizures, injunctions, or referral to the Department of Justice (DOJ) for civil penalties or criminal prosecution, including fines and for willful misbranding. For (Paxil), approved by the FDA in 1992 for adult and certain anxiety disorders but not pediatric use until later rejections, GlaxoSmithKline (GSK) began promoting it for adolescent in the early despite internal recognition that clinical trials, including Study 329 (completed in ), failed to meet endpoints and indicated heightened risks like suicidality. GSK sales representatives distributed materials citing the 2001 Journal of the American Academy of Child & Adolescent Psychiatry publication of Study 329, which portrayed as effective and well-tolerated in adolescents, to physicians for off-label prescribing, even as FDA supplemental applications for pediatric approval were pending and ultimately denied in 2003 due to inadequate evidence. This practice contravened FDA guidance requiring promotional claims to be consistent with substantial evidence from adequate, well-controlled studies and balanced with risks, as outlined in regulations predating the 2007 FDA Amendments Act expansions on off-label communications. The regulatory breaches culminated in DOJ charges against GSK for misbranding of Paxil from 1998 to 2003, alleging via sales aids, speaker programs, and detailing that omitted Study 329's negative raw data—such as non-significance on primary efficacy measures (e.g., Depression Rating Scale change scores) and emergent suicidality in 5.1% of paroxetine-treated adolescents versus 1.5% on placebo—while emphasizing selective positive secondary outcomes. In 2012, GSK entered a deferred prosecution agreement, pleading guilty to the charges and paying a $757 million criminal fine and forfeiture for Paxil misbranding as part of a $3 billion global , the largest healthcare resolution at the time, underscoring against systemic off-label schemes rather than isolated claims. No FDA pre-approval of Study 329-based promotions occurred, as manufacturers self-certify , with post-market surveillance revealing the discrepancies.

Regulatory and Investigative Responses

United Kingdom Inquiries

In 2003, the UK's Medicines and Healthcare products Regulatory Agency (MHRA), through its Committee on Safety of Medicines (CSM), initiated an urgent review of paroxetine (marketed as Seroxat) following a BBC Panorama broadcast on 13 May 2003 that disclosed internal GlaxoSmithKline (GSK) emails indicating suppressed data on suicidality risks in paediatric patients from trials including Study 329. GSK submitted full clinical trial datasets, prompting the CSM Expert Working Group to analyze efficacy and safety in children and adolescents. The review of three paroxetine trials (767 patients total, 378 on paroxetine at 10-50 mg/day for 8-12 weeks) found no demonstrated efficacy against major depressive disorder, with common adverse events including emotional lability, hostility, and insomnia leading to discontinuation in 10% of paroxetine-treated patients versus 5% on placebo. Suicide-related adverse events were reported in 3.7% (14/378) of recipients compared to 2.5% (7/285) on , yielding an of 1.5 (p=0.5), indicating an elevated though not statistically significant risk that, combined with the absence of efficacy, tipped the risk-benefit balance unfavorably. Research Database (GPRD) analyses further supported heightened non-fatal suicidal behavior risk with versus tricyclic antidepressants in under-19s (adjusted 1.6, 95% 1.0-2.5). On 10 June 2003, the MHRA issued guidance contraindicating for treating depressive illness in under-18s, noting its unlicensed status for this group despite prior in approximately 8,000 patients. The CSM's broader Expert Working Group report, finalized in 2004 after reviewing all SSRIs, affirmed paroxetine's contraindication and extended cautions to , sertraline, , , and for under-18s due to similar inefficacy and suicidality risks, exempting based on modest from two trials despite its own elevated events (10.2% versus 5.4% ). In December 2003, the MHRA advised against routine SSRI use in paediatric except , emphasizing close monitoring for , , and early in treatment. These actions stemmed from integrated of selective reporting in GSK submissions, as Study 329's published claims of and tolerability contrasted with showing neither.

United States Actions

![Eliot Spitzer, New York Attorney General who initiated legal action against GlaxoSmithKline over Paxil promotion][float-right] In June 2004, New York Attorney General Eliot Spitzer filed a civil lawsuit against GlaxoSmithKline (GSK) in New York State Supreme Court, alleging consumer fraud for concealing negative results from clinical trials, including Study 329, which demonstrated that paroxetine (Paxil) was ineffective for treating depression in children and adolescents and associated with increased risks of suicidality and other adverse events. The complaint highlighted GSK's promotion of Paxil for off-label pediatric use despite internal knowledge that the drug failed to show efficacy over placebo in Study 329, conducted from 1994 to 1998. The lawsuit was settled in August 2004, with GSK agreeing to post summaries of all results for its drugs on a public website within 30 days and to refrain from suppressing negative data in future promotions, without admitting liability; GSK also covered the state's legal costs estimated at around $250,000. The U.S. (FDA) reviewed data, including from Study 329, and in 2003 determined the drug was not proven safe or effective for treating in children and adolescents, leading to a rejection of GSK's supplemental application for pediatric labeling. In response to emerging safety signals, the FDA issued a advisory on June 19, 2003, of increased risk of suicidality in pediatric patients treated with Paxil, and by October 2004, mandated a on all antidepressants regarding heightened suicide risk in youth under 18. In July 2012, the U.S. Department of Justice (DOJ) announced a $3 billion settlement with —the largest healthcare fraud settlement in U.S. history—resolving civil and criminal allegations, including GSK's unlawful promotion of for pediatric use from 1998 to 2003 despite negative findings. pleaded guilty to misdemeanor misbranding under the Food, Drug, and Cosmetic Act for falsely representing as safe and effective for adolescents, specifically citing its role in authoring and approving the 2001 article that misrepresented 's results as demonstrating efficacy. The settlement included $2 billion in criminal fines and forfeiture, plus $1 billion in civil settlements for off-label promotion and failure to report safety data.

Outcomes of Investigations

In the United States, investigations into GlaxoSmithKline's (GSK) promotion of (Paxil) for adolescent , relying on selective interpretations of Study 329, led to multiple settlements. In August 2004, GSK settled a consumer fraud lawsuit filed by New York , agreeing to pay $2.5 million and to post summaries of all data, including negative results from Study 329, on its corporate website to enhance . This action addressed allegations that GSK had suppressed data showing paroxetine's lack of efficacy and potential harms in youth, which had fueled off-label prescribing. The Spitzer case contributed to expanded probes, culminating in a 2012 civil and criminal settlement with the U.S. Department of Justice (DOJ). GSK paid $3 billion—the largest healthcare fraud settlement in U.S. history at the time—including a $1 billion criminal fine and forfeiture, after pleading guilty to three counts, two of which involved misbranding Paxil and another drug with intent to defraud or mislead. The DOJ highlighted Study 329, alleging GSK falsely promoted it in publications and materials as demonstrating paroxetine's efficacy and safety for patients under 18, despite internal analyses showing otherwise on primary and secondary outcomes. An additional $14 million multi-state settlement in 2006 resolved related claims over Paxil's pediatric promotion. In the , the Medicines and Healthcare products Regulatory Agency (MHRA) investigated disclosures of unpublished adverse data from Study 329 and other trials in children, prompted by media reports and parliamentary inquiries in 2003. On December 10, 2003, the MHRA issued guidance advising against prescribing (Seroxat) to those under 18 due to evidence of doubled and behavior risks compared to , leading to the suspension of its pediatric license. This regulatory response extended to restricting other selective serotonin reuptake inhibitors except , influencing actions, though no financial penalties were levied on GSK. The outcomes emphasized data transparency reforms over punitive measures.

Civil Lawsuits

In June 2004, Attorney General filed a civil lawsuit against GlaxoSmithKline (GSK), accusing the company of consumer fraud for promoting Paxil () as safe and effective for treating in children and adolescents, despite internal knowledge from clinical trials, including Study 329, indicating a lack of efficacy and elevated risks of suicidality. The suit specifically charged GSK with suppressing negative results from four pediatric trials, misrepresenting data to physicians and the public through sales representatives, and distributing materials that falsely portrayed Paxil as beneficial for underage patients. The litigation prompted GSK to agree in August 2004 to publicly post raw data and clinical study reports from its pediatric Paxil trials on a , marking an early push for greater in pharmaceutical data. This initial resolution did not include monetary penalties but set the stage for broader scrutiny. In March 2006, Spitzer reached a multi-state with GSK valued at approximately $14 million, resolving allegations of fraudulent of Paxil for off-label pediatric use across multiple jurisdictions; the funds supported and on pediatric safety rather than direct restitution. Civil claims extended beyond state actions to federal proceedings under the , where whistleblowers and the U.S. Department of Justice pursued GSK for inducing false claims reimbursements through off-label promotion of Paxil supported by misrepresented Study 329 findings. These efforts contributed to GSK's $3 billion settlement in July 2012, of which over $2 billion addressed civil liabilities for fraudulent promotion practices, including the portrayal of Paxil as effective in adolescents contrary to trial evidence. GSK did not admit liability in these resolutions but ceased certain promotional activities and enhanced compliance measures. Private civil lawsuits from patients or families alleging harm from Paxil-induced suicidality in adolescents often referenced Study 329's concealed risks, though most were consolidated into multidistrict litigation focused on failure-to-warn claims rather than direct challenges to the study's publication. Outcomes varied, with GSK prevailing in several trials by arguing insufficient causation evidence, but the cumulative legal pressure underscored systemic issues in data disclosure tied to the study.

Criminal and Settlement Aspects

In July 2012, GlaxoSmithKline (GSK) agreed to plead guilty to three felony counts under the Food, Drug, and Cosmetic Act, including two counts of introducing misbranded drugs into interstate commerce—specifically Paxil () and Wellbutrin (bupropion)—through off-label promotion for pediatric without FDA approval. The plea resolved allegations that GSK promoted Paxil as safe and effective for children and adolescents despite internal knowledge of inadequate efficacy and elevated risks of suicidality from clinical trials, including Study 329 conducted from 1994 to 1998. This criminal resolution formed part of a broader $3 billion settlement with the U.S. Department of Justice (DOJ), the largest healthcare fraud settlement in U.S. history at the time, encompassing $757 million in criminal fines directly tied to Paxil and Wellbutrin misbranding. The DOJ charged that GSK's promotional tactics for Paxil in youth included distributing articles and sales aids that misrepresented trial data, such as portraying Study 329's null efficacy results as supportive of benefit while downplaying adverse events like , , and observed in 6.7% of paroxetine-treated adolescents versus 1.5% on . Tactics encompassed funding "ghostwritten" articles citing favorable interpretations of pediatric data, sponsoring programs to influence prescribers, and incentivizing sales representatives with bonuses for off-label pediatric prescriptions, contributing to over $1 billion in improper reimbursements. The settlement also addressed GSK's failure to report accurate safety data from Paxil trials to the FDA, including underreporting suicidal behaviors in pediatric populations. Beyond the federal criminal plea, the $3 billion total included civil settlements resolving False Claims Act violations, with approximately $2.24 billion for civil liabilities related to off-label promotion across multiple drugs, including Paxil's pediatric use. States received portions of the funds, such as California's share supporting Medicaid recovery, stemming from GSK's alleged pattern of suppressing negative Study 329 outcomes while aggressively marketing to offset adult sales declines post-1998 patent concerns. GSK entered a corporate integrity agreement with the Office of Inspector General, mandating enhanced compliance monitoring for five years, though the company maintained it did not admit liability for all civil claims. No individual executives faced criminal charges in connection with these events.

Reanalyses and Scientific Reexaminations

RIAT Reanalysis

The Restoring Invisible and Abandoned Trials (RIAT) initiative reanalyzed Study 329 using the complete dataset, including the , appendices, and over 77,000 pages of case report forms provided by GlaxoSmithKline via its SAS Solutions platform. Published in on 16 September 2015 by Le Noury and colleagues, this independent examination followed the trial's original 1994–1996 protocol, applying last observation carried forward and multiple imputation to address . Efficacy assessments revealed no significant benefits for or high-dose over in treating adolescent major . Mean reductions in Hamilton depression rating scale (HAM-D) total scores were comparable across groups, with no statistical difference (p=0.20).
TreatmentMean HAM-D Change (95% CI)
10.7 (9.1 to 12.3)
9.0 (7.4 to 10.6)
9.1 (7.5 to 10.7)
Responder rates, defined as at least a 50% HAM-D reduction, similarly showed no advantage for active treatments. Harms analysis indicated elevated risks with both drugs. Paroxetine linked to 11 serious adverse events versus 2 for placebo, including 5 suicidal or self-injurious behaviors compared to 1 in placebo. Total adverse events totaled 481 for paroxetine (70 severe or very severe), exceeding placebo counts, with heightened psychiatric events (103 versus 24). Imipramine showed additional cardiovascular effects. The reanalysis concluded that neither nor proved efficacious, while both increased harms beyond , highlighting discrepancies with the original 2001 report's claims of paroxetine's superiority and safety. Authors emphasized that full data access exposed underreporting of suicidal risks and selective outcome reporting in the initial publication.

Methodological Criticisms and Defenses

The original methodological approach in Study 329 involved a randomized, double-blind, -controlled comparing (20-40 mg/day), (titrated to 200-300 mg/day), and in 275 adolescents aged 12-18 with , using an 8-week acute phase followed by an optional continuation phase. Primary efficacy endpoints were change in Hamilton Depression Rating Scale (HAM-D) total score and (CGI) severity score, analyzed via last observation carried forward (LOCF) intent-to-treat with , a standard but bias-prone method given the 36% discontinuation rate in the group versus 23% on . Critics, particularly in the 2015 RIAT reanalysis by Le Noury et al., highlighted selective outcome reporting as a core flaw: the pre-specified primary endpoints failed to demonstrate superiority over (HAM-D change: -9.7 vs. -10.2, p=0.60; : no significant difference), yet the 2001 publication emphasized post-hoc secondary measures like K-SADS-L item responders (65% vs. 41%, p=0.02) and CGIC responders, without acknowledging the primaries' failure or multiplicity adjustments for multiple comparisons. This approach inflated apparent efficacy by shifting focus to exploratory analyses, potentially capitalizing on chance in a trial with modest power (n=189 completers). Adverse event reporting drew further scrutiny for inconsistent and underemphasis of risks; the original dichotomized events as severe/non-serious based on intensity rather than clinical narrative, omitting full /behavior details from clinical study reports (CSRs). Reanalysis, coding all events using full CSR data and standardized criteria, identified 11 harm-related discontinuations on versus one on , including five suicidal acts (versus one on ) and one completion, contradicting claims of comparable tolerability and highlighting suppression of events labeled as "behavioral syndrome" or "." High dosing also confounded harms, as it similarly increased adverse events without efficacy. Defenses from original authors and GSK representatives asserted that the multi-faceted —incorporating dimensional (HAM-D) and categorical (responder) measures—provided a clinically nuanced suitable for pediatric trials lacking established standards, where binary outcomes might overlook partial benefits in a heterogeneous population. They argued LOCF was appropriate per contemporaneous guidelines (e.g., FDA 1998 draft) for handling in short-term studies, and that responder analyses aligned with regulatory precedents for antidepressants. Responses to the RIAT critiqued its post-hoc harm recoding as non-protocol-driven and reliant on incomplete subsets (e.g., excluding some non-random completers), potentially introducing bias, while maintaining the original's overall design met CONSORT-equivalent rigor for 1990s multicenter trials. However, these arguments have been undermined by FDA's 2002 rejection of pediatric approval, citing insufficient efficacy and emergent suicidality signals across trials, including Study 329.

Retraction Calls and Journal Responses

Calls for retraction of the Study 329 article, published in the Journal of the American Academy of Child and Adolescent Psychiatry (JAACAP) in 2001, emerged shortly after its release, with initial complaints directed to the journal in 2003 citing discrepancies between the reported findings and the underlying data. In 2005, philosopher and medical ethicist Leemon McHenry formally complained to then-editor Mina Dulcan, highlighting evidence of ghostwriting and selective reporting by GlaxoSmithKline (GSK) that misrepresented paroxetine's efficacy and safety in adolescents. These efforts intensified in December 2009 when Jon Jureidini and McHenry explicitly requested retraction, arguing that the paper conflated primary and secondary outcomes to fabricate positive results, violating standards of scientific integrity. JAACAP's leadership consistently rebuffed these demands, maintaining that the article's tables disclosed negative primary endpoint results sufficiently to preclude retraction. Editor-in-chief stated in responses to critics that no justification existed for retraction, emphasizing the paper's overall despite acknowledged limitations in . Following the 2015 RIAT reanalysis in , which re-examined raw data and concluded lacked efficacy while increasing harms like suicidality, additional calls for retraction were lodged, supported by a dedicated site (study329.org) tracking the campaign. The journal and the American Academy of Child and Adolescent Psychiatry (AACAP) upheld their position, with authors including B. Keller issuing a collective denial of ghostwriting allegations in a 2015 letter. Persistent refusal prompted legal action on September 8, 2025, when attorney George W. Murgatroyd III filed suit in D.C. Superior Court against AACAP and publisher , alleging the unretracted paper enabled fraudulent promotion of , contributing to adolescent harms including suicides in represented cases. In response, JAACAP issued an expression of concern on September 30, 2025—the first formal editorial notice after over two decades—stating it aimed "to alert readers to concerns that have been raised about the article" while an ongoing review proceeds, without conceding grounds for full retraction. Critics, including Jureidini, viewed this as a belated insufficient to rectify the paper's influence, which has garnered over 450 citations.

Broader Scientific Implications

Antidepressants Efficacy in Adolescents

The original publication of Study 329 in 2001 claimed that demonstrated efficacy in treating (MDD) in adolescents aged 12-18, reporting a response rate of 66% compared to 49% for based on secondary outcomes like the Clinical Global Impression-Improvement (CGI-I) scale. However, the trial's predefined primary outcomes—change in (HAM-D) total score and HAM-D-defined depressive response—showed no statistically significant differences between paroxetine and (p=0.11 for HAM-D change; response rates 60.3% vs. 56.9%). The 2015 RIAT reanalysis of the full dataset confirmed the absence of efficacy, finding inferior to on one primary outcome (HAM-D response: 56.9% vs. 60.3% , but not significant after multiplicity adjustment) and no benefit on the other, with also ineffective versus . This reexamination highlighted selective reporting in the original paper, where efficacy claims relied on post-hoc analyses of non-primary measures, contributing to overstated benefits amid high response rates typical in adolescent trials (often 40-50%). Broader meta-analyses of randomized controlled trials (RCTs) for selective serotonin reuptake inhibitors (SSRIs) in youth MDD reinforce limited efficacy signals. A 2016 network meta-analysis of 34 RCTs involving over 5,000 children and adolescents found only fluoxetine superior to placebo (standardized mean difference [SMD] -0.51, 95% credible interval -0.93 to -0.11), while paroxetine showed no efficacy advantage (SMD -0.07, -0.62 to 0.46) and higher adverse event-related dropouts.30385-3/fulltext) Other SSRIs like sertraline and citalopram similarly lacked robust evidence, with overall effect sizes small (SMD ~0.2-0.3 where positive) and confounded by publication bias favoring positive results.30385-3/fulltext) High placebo response rates—median 45% in pediatric RCTs, exceeding adult rates—further diminish apparent drug-placebo differences, often rendering active treatments statistically non-significant in low-response subgroups.30385-3/fulltext) Industry-sponsored trials, which dominate the evidence base, exhibit SMDs of 0.13-0.20 for antidepressants versus placebo in adolescents, but these shrink or reverse when excluding high-placebo-response studies or adjusting for selective outcome reporting. Regulatory approvals remain restricted: the U.S. FDA has greenlit only fluoxetine (2002) and escitalopram (2009) for pediatric MDD based on pivotal trials like TADS (fluoxetine SMD -0.68 vs. placebo), while paroxetine's application was rejected in 2003 due to inefficacy and harm signals from Study 329 and others. These findings underscore methodological challenges, including unblinding from side effects and overreliance on subjective scales, questioning causal antidepressant effects beyond nonspecific factors in this population.

Suicidality Risks: Evidence and Debates

In the original (CSR) for Study 329, conducted from 1994 to 1998, the group (n=93 in the acute phase) experienced 10 possibly suicidal events, compared to 1 event in the group (n=87). These included suicide attempts, gestures, and ideation, with 5 explicit suicide attempts reported in the arm during acute treatment versus none in . The 2001 publication in the Journal of the American Academy of Child & Adolescent Psychiatry minimized these risks, stating no clinically significant differences in adverse events related to suicidality between groups, attributing events to the underlying rather than the drug. The 2015 RIAT reanalysis of the full dataset, published in , confirmed elevated harms, including and behavior, in the group across acute, continuation, and taper phases: 23 suicidality-related events occurred in 15 patients versus 7 events in 5 patients. This reexamination classified events using standardized criteria, revealing that was associated with a higher incidence of serious adverse events (SAEs) involving or suicidality (11% of patients vs. 2% ), contributing to the conclusion of net harm without efficacy benefits. Continuation phase data further indicated 1 suicidal gesture or trauma event potentially linked to , alongside ongoing risks during taper. These findings informed the U.S. Food and Drug Administration's (FDA) 2004 black-box warning for all antidepressants, including , highlighting a doubled risk of suicidality (ideation, attempts, or behavior) in pediatric patients during initial treatment months, based on pooled analyses of 24 trials showing rates of 4% versus 2% for (risk ratio ≈2.0). specifically faced non-approval for adolescent in 2003 due to inefficacy and adverse signals, including from Study 329. Debates persist on causality and interpretation. Proponents of antidepressant use, including some industry-linked analyses, argue that elevated events reflect behavioral activation in severely depressed youth rather than direct causation, noting high baseline suicide risk in major depressive disorder and no completed suicides in the trial. Critics, drawing from raw CSR data and meta-analyses, contend this understates a class effect of selective serotonin reuptake inhibitors (SSRIs) like paroxetine, which may disinhibit impulses or induce akathisia-like states exacerbating suicidality, as evidenced by dose-dependent increases in events. Post-warning studies have debated net population effects, with some observational data suggesting reduced youth prescriptions correlated with stable or slightly rising suicide rates, potentially indicating undertreatment harms outweighing drug risks; however, randomized trial evidence, including Study 329, prioritizes the acute elevation in events as a causal signal requiring vigilant monitoring.

Transparency and Trial Data Access Lessons

The original publication of Study 329 in the Journal of the American Academy of Child & Adolescent Psychiatry in 2001 selectively reported outcomes, emphasizing efficacy and tolerability of while downplaying harms, including and behavior, as internal GlaxoSmithKline (GSK) documents later revealed through litigation. This selective reporting was facilitated by ghostwriting, where GSK employees drafted the manuscript and academic authors provided limited input, obscuring discrepancies between raw data and published claims. Access to full clinical study reports (CSRs), protocols, and individual patient-level data was restricted to GSK, preventing independent verification for over a decade and allowing the misleading conclusions to influence prescribing practices. The Restoring Invisible and Abandoned Trials (RIAT) initiative reanalysis, published in on September 16, 2015, overcame these barriers by compiling data from publicly available FDA review documents, CSRs obtained via legal discovery in U.S. lawsuits against GSK, and other regulatory filings, rather than relying on sponsor-provided summaries. This effort demonstrated that lacked efficacy for adolescent major depression on primary outcomes and was associated with significantly higher rates of harms, including suicidality ( 4.63 for or behavior versus ), contradicting the original report's narrative. The process highlighted practical challenges, such as reconciling inconsistent datasets across sources and the absence of a complete , underscoring how impedes rigorous scrutiny. Key lessons from Study 329 emphasize the necessity of mandatory, prospective access to raw trial data, protocols, and CSRs to enable independent reanalyses and mitigate selective reporting. Transparent reporting standards, such as those advocated by RIAT, reveal that inadequate design flaws often underlie poor outcomes, as evidenced by the trial's underpowered sample and flexible efficacy endpoints not predefined in the protocol. Open databases for individual participant data would facilitate network meta-analyses and harm assessments, reducing reliance on sponsor interpretations prone to commercial bias, while pre-registration of trials and outcomes could prevent post-hoc adjustments. These reforms address systemic issues where pharmaceutical companies historically withheld negative results, as GSK did here, contributing to distorted evidence bases in . Broader implications include advocacy for enforceable data-sharing mandates, as seen in post-Study 329 pushes for initiatives like the European Medicines Agency's clinical data and the AllTrials , though implementation gaps persist, particularly for harms data in older trials. Without such , public health risks from ineffective or harmful drugs endure, as paroxetine's adolescent approval pathway was influenced by this and similar misrepresented studies until regulatory reevaluations in the mid-2000s. The case exemplifies how litigation, rather than routine scientific processes, often forces , reinforcing calls for independent data custodians to prioritize evidentiary integrity over proprietary interests.