Part I. Executive Summary & the Placebo Effect and Studying Antidepressant Efficacy 2018-02-07T22:20:46+00:00

Part I. Executive Summary

Part I | Part II | Part III | Part IV | Part V | Part VI

In 2010, the USAO in the District of Massachusetts entered into a series of agreements with Forest Laboratories, Inc. and Forest Pharmaceuticals, Inc., (“Forest”) relating, in part, to the off-label (unapproved) promotion of Celexa (citalopram) and Lexapro (escitalopram) for use in children and adolescents.[1]  Forest agreed to plead guilty to engaging in a misdemeanor count of off-label promotion for Celexa between 1998 and 2002 and pay $39 million in fines.[2]  Additionally, Forest entered into a civil settlement agreement to resolve, in part, allegations that Forest fraudulently induced false claims for the pediatric use of Celexa and Lexapro to be submitted to government healthcare payers between 1998 and 2005.[3]  Forest agreed to pay $149 million to settle these claims.[4]  Forest also entered into a Corporate Integrity Agreement (“CIA”) designed to monitor the promotional practices of Forest for a period of five years.[5]  The Plea Agreement, Civil Settlement, and CIA were conditioned on each other[6] and required that Forest be honest with the USAO about its conduct.[7]

A central feature of Forest’s wrongful conduct involved the promotion and dissemination of a “positive” Celexa double-blind, placebo-controlled clinical trial in children and adolescents, MD-18, and the suppression of a negative Celexa double-blind, placebo-controlled clinical trial in adolescents, Study 94404.  Forest’s one-sided presentation of the efficacy data raised concerns about how companies such as Forest disclose and use data collected during clinical trials, particularly when used as part of an off-label promotion campaign.  Indeed, the factual claim of one “positive” trial and one “negative” trial played an important role in the USAO’s prosecution of the original case against Forest.

Recently unsealed documents and testimony, however, show that Celexa Study MD-18 was not a “positive study” and that Forest misled the FDA, the USAO, and the public about this fact.  In other words, a material fact that formed the basis of the USAO’s and Forest’s negotiations was, at that time, false, and Forest knew it.  Moreover, this misconduct does not stop there.  Shortly before the USAO and Forest finalized their agreements, the FDA approved Lexapro for use in adolescents, based in part on the misrepresented MD-18 (Celexa) study.  The fact that Forest obtained FDA approval for Lexapro for use in adolescents militated against the government’s prosecution.  The 2009 Lexapro approval, however, was based on the false claim that MD-18 was positive—a false assertion made to the FDA in 2002 (and reasserted to the FDA in 2008 as part of Forest’s supplemental New Drug Application (“sNDA”) for Lexapro).  If MD-18 had properly been disclosed as negative, the FDA would not have approved Lexapro for use in adolescents[8] and the government’s prosecution of Forest would have included Forest’s misrepresentations regarding Celexa’s efficacy in two studies—not just the suppression of Study 94404.

The issue centers on how Forest manipulated the MD-18 data to obtain a “positive” result.  All of the secondary endpoints for MD-18 were negative, meaning Celexa did not outperform placebo in treating depression on all four of the pre-specified secondary endpoints.[9]  Moreover, of those patients who completed the study, i.e., the observed cases, there was also no statistical difference between Celexa and placebo.[10]  However, Forest represented to the FDA, USAO, and others that the primary endpoint for MD-18 was positive because, although the difference was very small, Celexa appeared to outperform placebo to a statistically significant degree.  It turns out, however, that this “positive” result was based on data from nine patients who were unblinded during the study.  When the data from these unblinded patients is removed, however, the primary result is negative—indeed, the results are negative across the board on every primary and secondary endpoint.[11]

How did this happen?  At the beginning of the clinical trial, two clinical investigators informed Forest that some of their patients were receiving pink pills and others were receiving white pills.[12]  This prompted an investigation by Forest, which discovered that a packaging error had caused the medication for the patients randomized to the Celexa group to be pink, Forest-stamped, dose-stamped, oval-shaped, commercial Celexa tablets.[13]  Forest immediately notified the clinical investigators that the pink pills were commercial Celexa and instructed them to replace the medication with properly blinded white pills.[14]  However, for the nine patients already randomized to the trial, Forest directed the clinical investigators to continue the patients with the wrong-colored pills.[15]  This meant the first nine patients were unblinded—the investigators knew the patients taking the pink pills were taking Celexa and the patients taking white pills were on placebo.

In March 2000, after Forest learned of the dispensing error but before the results of the study were known, Forest drafted a letter to send to the FDA explaining the situation.  In the original draft, Forest stated that the dispensing error could have “unblinded the study.”[16]  However, Forest’s regulatory affairs manager, Amy Rubin, changed the letter to state that the dispensing error had the “potential to cause patient bias.”[17]  This prompted Forest’s medical director, Dr. Charles Flicker, to respond: “Altho ‘potential to cause bias’ is a masterful stroke of euphemism, I would be a little more up front about the fact that the integrity of the blind was unmistakenly violated.”[18]  Un-phased by Dr. Flicker’s concern that Forest was not being “up front” with the FDA, Ms. Rubin responded: “Thanks for the compliement [sic].  Part of my job is to create ‘masterful’ euphemisms to protect Medical and Marketing.”[19]  For Ms. Rubin, misleading the FDA was not only acceptable, it was part of her job.  And, she did her job well.  The letter sent to the FDA in March 2000 used the masterful euphemism language.[20]

The letter did, however, state that “[f]or reporting purposes, the primary efficacy analysis will exclude the . . . potentially unblinded patients[.]”[21]  Thus, before Forest had the results of MD-18, Forest recognized the data was corrupted and promised that “[a] full complement of 160 patients” would still be “enrolled under standard double-blind conditions.”[22]

The results for MD-18 were revealed to Forest in August 2001 and Forest learned, for the first time, that if the nine unblinded patients were excluded from the analysis, as it had promised the FDA it would do, the results would be negative.[23]  But, Forest reneged on its promise to the FDA.  When it submitted the final study report to the FDA in April 2002, Forest included the unblinded patients in the primary efficacy analysis and buried, in an appendix, the results of the primary efficacy analysis excluding the unblinded patients.[24]  In the narrative section of the report, Forest explained that there had been a dispensing error where nine patients received pink-colored pills, but the patients “were otherwise blinded.”[25]  This is in stark contrast to Dr. Flicker’s unequivocal pronouncement that the integrity of the blind was unmistakenly violated.  In its submission to the FDA, Forest did not disclose that the investigators were unblinded or that the medication dispensed was Forest-branded, dose-stamped, oval-shaped commercial Celexa tablets.  When the FDA reviewed the results of MD-18, it copied and pasted the language from the final study report and parroted the claim that pink pills were dispensed but were “otherwise blinded.”[26]  Forest’s deception worked—the FDA had no idea that the nine patients were actually unblinded[27] and that the study, when properly analyzed, was negative across the board.

Before the MD-18 study report was even written or given to the FDA, Forest started promoting the “positive” results of MD-18 to physicians.  Forest issued a press release emphasizing the importance of the positive MD-18 study in a field, i.e., SSRI treatment of pediatric depression, which had consistently failed to produce positive results[28]; paid Dr. Karen Wagner (an investigator on the study) to present the false “positive” results of the study at various academic conferences and, directly, to physicians in CME programs and in-person off-label promotion meetings[29]; and published the false results of MD-18 in a ghostwritten manuscript and then instructed its sales force to use the publication to promote the use of Celexa and Lexapro in children.[30]  None of these presentations and publications disclosed the unblinding issue.  It was buried.

The impact of the off-label promotion of the false data was known to Forest, as demonstrated in the following slide taken from Forest’s internal marketing plan[31] discussing its anticipated launch of Lexapro:

When Forest’s off-label promotion was finally exposed by the USAO in 2010 and Forest was forced to settle and plead guilty to the crime, Forest did not disclose its fraud related to MD-18.  Instead, Forest represented to the USAO and DOJ that MD-18 was positive, militating its misconduct in suppressing the dissemination of Study 94404.  Thus, Forest’s false assertion that MD-18 was a positive study formed, in part, the basis of the USAO’s negotiated settlements with Forest.  These documents and testimony clearly demonstrate that Forest made material misrepresentations to the USAO and FDA about this issue and, in fact, continues to do so to this very day.  When Dr. William Heydorn was deposed, the former Forest scientist responsible for preparing the MD-18 final study report and a named author on MD-18’s publication with Dr. Wagner, he admitted “I wish we had done things a little differently . . . probably should have been more forthcoming[.]”[32]

PART I:  The Placebo Effect And Studying Antidepressant Efficacy

All drugs are susceptible to the placebo effect—the effect a drug has on a patient that has nothing to do with the medicinal properties of the drug but is caused by the very act of getting medical attention.[33] The belief that one is possibly experiencing medical treatment, by itself, can create significant and measurable improvement for many conditions.[34]

In 1962, reeling from news of birth defects caused by a drug called thalidomide, Congress amended the Food Drug and Cosmetic Act (the Kefauver Harris Amendment, Pub. L. No. 87-781, 76 Stat. 780 (1962)).  Before a drug could be sold as an effective medication, the drug maker would be required to prove the drug could outperform placebo or, in other words, demonstrate that the benefit patients receive from a drug could not simply be duplicated by administration of placebo.[35]

Today, a drug’s efficacy is determined using double-blind randomized controlled trials (“DBRCTs”).[36]  A DBRCT involves the systematic comparison of patients taking a drug and patients taking a placebo.[37]  Patients enrolled in the clinical trials are randomly assigned into two groups.  One group takes the drug and the other takes a placebo.  However, neither the investigators nor the patient know which group each patient is in.  Once the study is complete, the benefit observed in the two groups is compared, and if the patients taking the drug meaningfully outperform the patients in the placebo group, the clinical trial is considered positive.[38]  If the drug does not outperform placebo, it is called negative.

As its name suggests, a DBRCT involves three elements, all of which are designed to limit bias: (1) double-blind (2) randomized (3) controlled trials.[39]  First, the trial must be double-blind.  This means neither the investigator nor the patient know whether the pill ingested by the patient is the active drug or placebo.[40]  If either the investigator or the patient is unblinded, it invalidates the data since there is no way to determine whether the effects observed are caused by the drug or caused by the placebo effect (for the patient and investigator).  Second, the trial must be randomized.[41]  Patients assigned to the drug or control group must be randomly assigned.  Otherwise, the distribution of patients would, itself, inject bias into the study.  Finally, the trial must be controlled.  This means the drug must be compared to a control group, i.e., a placebo pill.[42]

Before a DBRCT is conducted, a study protocol is generated.[43]  The protocol specifies the study’s endpoints—the primary and secondary measures that determine whether the drug works—and the conduct / procedures of the study.  In nearly all DBRCTs, before a study will be considered positive, the primary endpoint must statistically outperform placebo.  This means that the difference between the drug and placebo must be large enough to conclude the difference was not a result of chance.  Conventionally, and for the purposes of the DBRCTs discussed in this memorandum, to be considered statistically significant, the endpoint must have a p-value (a statistical measure) less than 0.05.[44]

Part I | Part II | Part III | Part IV | Part V | Part VI


[1] See Exh. 1, DOJ Press Release.

[2] Exh. 2, Plea Agreement at 5; see Exh. 3, Criminal Information ¶¶ 55-71 (outlining allegations); Exh. 5, Side Letter Agreement with Forest Laboratories.

[3] Exh. 4, Civil Settlement Agreement at pp.3-4, ¶¶ G(1)-G(3).  While the settlement encompassed the years 1998-2005, there is evidence that Forest’s sales representatives were illegally off-label promoting Celexa and Lexapro through 2009. One of Forest’s marketing executives testified: “I have knowledge that representatives may have presented Celexa or Lexapro inappropriately” and, when asked, “Between 2002 and 2009?” the marketing executive replied: “Yes.”  Exh. 7 Azari Depo at 236:1-237:22.

[4] Id. at pp. 6, ¶ 1.

[5] Exh. 6, Corporate Integrity Agreement.

[6] Exh. 2, Plea Agreement at pp. 11; Exh. 4, Civil Settlement Agreement at pp.3 ¶ E, pp. 10 ¶ 5; Exh. 6, Corporate Integrity Agreement at 1.

[7] Exh. 2, Plea Agreement at pp. 6; Exh. 4, Civil Settlement Agreement at pp.17 ¶ 15.

[8] See Exh. 8, 2017 Depo. of T. Laughren at 401:15-402:10 (Dr. Thomas Laughren, the senior FDA official who approved Lexapro for use in adolescents admitting that he would not have approved Lexapro for adolescents if MD-18 was negative).[9] Exh. 9, Excerpts of Study MD-18 Rpt. at pgs. 101-104, 244.[10] Id. at 111 (listing p-value of observed cases analysis at week at as 0.1670); Exh. 8, 2017 Depo. of T. Laughren at 97:1-21, 99:18-21, 343:6-10 (“Q. Sure.  But we know that the OC results for the people who actually completed the clinical trial, that actually was negative for efficacy, right? A. That’s true.”); Exh. 11, 2016 Depo. of W. Heydorn at 138:24-139:6, 144:6-9.[11] Exh. 12, 2016 Depo. of S. Closter (Forest’s Rule 30(b)(6) Corporate Representative) at 294:10-295:20 (“If they were removed from the study, I understand that the result would have been negative.”).[12] See Exh. 13, Draft FDA Letter with C. Flicker Handwritten Comments at 1 (describing how Forest learned of the dispensing error).[13] Id.; see Exh. 14, Memo re. Investigation of CIT-MD-18 Clinical Study Use of Trade-Dress Citalopram 20 mgs Tabs at 1-2; Exh. 15, Memo re. CIT-MD-18 (Deviation Report) at 1-2 (“It was brought to our attention  . . . some patients enrolled in this study had pink tablets in their bottles. We immediately investigated . . . We discovered . . . the pink oval tablets with FP/20MG imprints.”).[14] Exh. 16, Email re. CIT-18 FAX to Investigational sites (w/ attachment) at 1 (“[A] copy of the FAX that went out to all CIT-MD-18 Pediatric Investigational sites this morning is attached[.]”).[15] Id. at 2 (directing patients already randomized to continue on with study).[16] Exh. 17, Email re. Letter to FDA for CIT-18 (w/attachment) at 2 (“The purpose of this letter is to inform the agency that due to a clinical supplies packaging error for the above-referenced trial, eight randomized patients at two investigational sites were dispensed medication that could have potentially unblinded the study.”).[17] Exh. 18, Email responses re. Letter to FDA for CIT-18 at 1.[18] Id.[19] Id.[20] Exh. 19, Letter from T. Varner (Forest) to R. Katz (FDA) at 1.[21] Id.[22] Id.[23] See Exh. 20, Email re. CIT-MD-18 at 1 (“We need to generate Tables 4.1A and 4.1B for ITT population, excluding the 9 patients who were unblinded at the beginning of the study. Can you please tell Qiong who they are and try to get the results before 9:30, Friday morning?”).[24] See Exh. 21, Email re. Notes from conference call Oct 4 (w/attachment) at 2 (“[S]ome citalopram table[t]s were not blinded. The 9 patients who received unblinded medication were included in the main analyses; a secondary ‘Post-hoc analysis of the ITT subpopulation’ was done. Refer to these analyses briefly in methods and results and reference the reader to the appendix table.”); Exh. 9, Excerpts of Study MD-18 Rpt. at pgs. 70, 244 (unblinded results).[25] Id. at pg. 44.[26] Exh. 22, Review and Evaluation of Clinical Data by Dr. Earl Hearst, FDA at 11.  All but two words of Dr. Hearst’s medical review of MD-18 were copied and pasted from the final study report.[27] Exh. 8, 2017 Depo. of T. Laughren at 154:6-23 (“18 Q Okay. So it was your understanding that the patients, despite receiving different color tablets, were still blinded, correct? . . . THE WITNESS: Well, that — that was — that was my assumption, correct.”).[28] Exh. 23, 2001 Forest Press Release at 2-3 (“This study is significant because few studies involving any antidepressant have shown efficacy compared to placebo in the treatment of depression in children and adolescents . . . Citalopram is now one of the few therapies for which we have data showing safety and efficacy for this population.” (quoting Dr. Karen Wagner)).[29] Exh. 24, Email re. Ped data at 2 (“[W]e would like to wrap some PR and CME around this data”); Exh. 25, Email re. ACNP pediatrics abstract at 1 (“John wants GCI to start working a release and any other way they can spin this data.”); Exh. 26, Emails re. ACCAP meeting at 1 (“You should discuss with GCI bringing her [Dr. Wagner] in for media training prior to the start of the CME program.”); Exh. 27, Emails re. ACCAP Meeting at 3 (“We spoke with Karen Wagner today about the current state of affairs regarding the pediatric data. . .  She . . . reminded us that if we want to appeal to the PCP and Pediatric audiences, we need to publish in a place that provided the appropriate readership . . . She also said that the lack of data regarding the use of Celexa for pediatrics is limiting it to ‘last choice’ among physicians – she just wanted to make sure we understood the marketing advantages of the data.”).[30] See, e.g., Exh. 28, Selection of Call Notes at 7, 16-17 (“discussed cx used in children . . . and results of dr wagner study regarding cx use for children and adolescents . . . Brought up the Wagner study and sent study to Dr. asked Dr if it would make a difference to use Lx in that age group since Cx has done well.”).[31] Exh. 29, Lexapro Tactical Presentation at pg. 12; see also id. at pgs. 10-14 (discussing strategies to increase under 20 market).[32] Exh. 11, 2016 Depo. of W. Heydorn at 307:24-308:15.[33] Exh. 30, U.S. Food and Drug Administration (FDA), Guidance for Industry, E 10 Choice of Control Group and Related Issues in Clinical Trials, at 4 (May 2001).[34] Id.[35] See 21 C.F.R. § 314.126.[36] In re Neurontin Mktg. & Sales Practices Litig., 712 F.3d 21, 47-49 (1st Cir. 2013).[37] See FDA, supra note 33, at 4-5.[38] Id. at 5.[39]  Exh. 31, Food and Drug Administration (FDA), Guidance for Industry, E9 Statistical Principles for Clinical Trials, at 10-14 (Sept. 1998).[40] Id. at 10 (“Blinding or masking is intended to limit the occurrence of conscious and unconscious bias in the conduct and interpretation of a clinical trial arising from the influence that the knowledge of treatment may have on the recruitment and allocation of subjects, their subsequent care, the attitudes of subjects to the treatments, the assessment of end-points, the handling of withdrawals, the exclusion of data from analysis, and so on.”); see FDA, supra note 33, at 4 (listing possible ways bias enters a trial without blinding).  In the context of clinical trials related to depression, this factor is particularly important where a patient’s depression is assessed by an investigator based on the patient’s answers to specified questions about how they feel.  If either the investigator or the patient knows they are receiving the drug, that knowledge will likely influence their assessment.[41] FDA, supra note 39, at 12 (“In combination with blinding, randomization helps to avoid possible bias in the selection and allocation of subjects arising from the predictability of treatment assignments.”).[42] Id. at 18.[43] Id. at 3 (“For each clinical trial contributing to a marketing application, all important details of its design and conduct and the principal features of its proposed statistical analysis should be clearly specified in a protocol written before the trial begins.”).  The protocol must be followed religiously. See Exh. 32, Ravindra B. Ghooi, et al., Assessment and classification of protocol deviations, 7 Perspectives Clin. Res. 3, 132-36 (July-Sept. 2016) (discussing importance of following protocols and how deviating from them can lead to misleading study results); Exh. 33, Stephen L. George & Marc Buyse, Data fraud in clinical trials, 5 Clin. Invest, 2 161-173 (2015) (discussing how falsification of data, i.e., misrepresenting important events in a clinical trial, are the most egregious types of misconduct).[44]  Exh. 34, Food and Drug Administration (FDA), DRAFT Guidance for Industry, Multiple Endpoints in Clinical Trials, at 4-5 (Jan. 2017) (discussing the typical use of a p-value of less than 0.05).