Aug 15, 2004 - Mosholder Report

[Guest guest]

> SSRI-Research

> Sat, 21 Aug 2004 17:38:16 -0400

> [sSRI-Research] Aug 15, 2004 - Mosholder

> Report

>

> M E M O R A N D U M

>

> DEPARTMENT OF HEALTH AND HUMAN SERVICES

> PUBLIC HEALTH SERVICE

> FOOD AND DRUG ADMINISTRATION

> CENTER FOR DRUG EVALUATION AND RESEARCH

> PID# D040495

>

>

http://www.fda.gov/ohrms/dockets/ac/04/briefing/2004-4065b1-11-TAB09a-Mosholder-\

review.pdf

>

> DATE: August 16, 2004

>

> Paul Seligman, M.D., M.P.H.

> Acting Director, Office of Drug Safety, HFD-400

> (hard copy signed 8-16-04)

> Anne Trontell, M.D., M.P.H., Deputy Director

> Office of Drug Safety, HFD-400

> (hard copy signed 8-16-04)

>

> TO: Russell Katz, M.D., Director

> Division of Neuropharmacological Drug Products,

> HFD-120

>

> SUBJECT: Office of Drug Safety Cover Memorandum

>

> Follow-up Consult of August 16, 2004 by Andrew

> Mosholder on

> Suicidality in pediatric clinical trials with

> paroxetine and other

> antidepressant drugs:

>

> Drugs: paroxetine, sertraline, venlafaxine,

> fluoxetine, fluvoxamine,

> citalopram, nefazodone, mirtazapine, and bupropion

>

> The results of Dr. Mosholder's analyses (dated

> February 18, 2004) are very similar to those

> obtained using other statistical methods and a

> reclassification of suicidality events by Columbia

> University. Remaining questions to be addressed are

> whether the overall finding of increased risk

> applies to all or selected drug products among the

> nine products studied, and what additional

> regulatory actions are merited. On those topics, we

> reference the Office of Drug Safety memorandum

> written by Anne Trontell and dated March 15, 2004.

>

>

>

> M E M O R A N D U M

>

> DEPARTMENT OF HEALTH AND HUMAN SERVICES

> PUBLIC HEALTH SERVICE

> FOOD AND DRUG ADMINISTRATION

> CENTER FOR DRUG EVALUATION AND RESEARCH

> PID# D040495

> DATE: August 16, 2004

>

> FROM: Andrew D. Mosholder, M.D., M.P.H.,

> Epidemiologist

> Division of Drug Risk Evaluation, HFD-430

>

> THROUGH: Mark Avigan, M.D., C.M., Director

> Division of Drug Risk Evaluation, HFD-430

> (hard copy signed 8-16-04)

>

> TO: Paul J. Seligman, M.D., M.P.H., Acting Director

> Office of Drug Safety, HFD-400

> Anne Trontell, M.D., M.P.H., Deputy Director

> Office of Drug Safety, HFD-400

>

> SUBJECT: Suicidality in pediatric clinical trials of

> antidepressant drugs:

> Comparison between previous analyses and Colombia

> University

> classification

>

> Drugs: paroxetine, sertraline, venlafaxine,

> fluoxetine, fluvoxamine,

> citalopram, nefazodone, and mirtazapine

>

> BACKGROUND

>

> Please refer to the 3-19-04 consult1 regarding

> suicidal adverse events in pediatric clinical trials

> of antidepressant drugs. That consult described a

> meta-analysis by the undersigned of suicidal adverse

> events in short-term placebo-controlled pediatric

> clinical trials, showing a statistically

> significant association of suicidal adverse events

> with antidepressant drug treatment. However, because

> of concerns regarding misclassification of cases,

> FDA expanded the case finding algorithm, and

> arranged to have expert consultants jury the cases

> prior to any definitive analyses.

>

> Please refer to the materials from the February 2,

> 2004 Advisory Committee Meeting on this topic for

> additional details.

>

> The aforementioned case reclassification has

> recently been completed by an expert panel convened

> by Columbia University. Dr. Tarek Hammad of FDA's

> Division of Neuropharmacological Drug Products

> (DNDP) has performed a new meta-analysis, based on

> the reclassification that was performed by Columbia

> University, which I will refer to herein as the DNDP

> analysis. I was asked to examine the impact of the

> Columbia University reclassification of cases on my

> analysis performed prior to the Columbia University

> reclassification, as described in 1 PID# D030341 the

> 3-19-04 consult, which I will refer to in this

> memorandum as the ODS analysis. I have also compared

> my results, that were obtained using different

> analytic methods, with those of Dr. Hammad. For

> these purposes, Dr. Hammad has kindly provided me

> with his results, which are

> included below.

>

> METHODS

>

> A full description of the methodology is beyond the

> scope of this memorandum, so the interested reader

> should refer to the reviews by Dr. Hammad and myself

> for details on the analytic methods and clinical

> trial data. It should be noted that the two analyses

> used different case ascertainment strategies and

> different case criteria; the ODS analysis used only

> cases identified by the sponsors through an

> electronic search of their adverse event databases,

> while the DNDP analysis supplemented this approach

> with additional search methods. Some salient

> differences between the two analyses are the

> following: (1) the ODS analysis did not employ a

> correction for zero cells, whereas the DNDP analysis

> does; (2) the ODS analysis used rate ratios, with

> person-time for

> denominators, while the DNDP analysis uses risk

> ratios, with numbers of patients for denominators;

> (3) the ODS analysis included events occurring up to

> 30 days after discontinuation of treatment, while

> the DNDP analysis uses a 1-day post-treatment

> window; and (4) the ODS analysis included taper

> phase events, which are excluded from the DNDP

> analysis.

>

> This memorandum will present the following three

> modes of comparing the DNDP and ODS analyses.

>

> Comparison of risk ratios obtained with ODS and DNDP

> statistical methods

> In order to determine how the two different

> statistical methods affect the values for the

> relative risks, we compare the relative risks

> obtained with identical patient populations and

> classifications of cases. For this purpose, all

> possibly suicide-related events were included

> regardless of whether

> they were serious or not; this outcome variable was

> common to both sets of data, thereby allowing a

> comparison. While this outcome may be accorded

> relatively little inferential value because of the

> concerns about case misclassification, it does

> permit a direct comparison between results from the

> two statistical methods.

>

> Comparison of case classifications

>

> In order to assess the degree of agreement or lack

> thereof between the Columbia University and ODS case

> classifications, the " primary " outcome for the

> respective analyses must be defined. For the DNDP

> analysis, this is " Outcome 3, " definitive suicidal

> behavior/ideation, a composite of Columbia

> University codes 1 (suicide attempt), 2 (preparatory

> actions towards imminent suicidal

> behavior), and 6 (suicidal ideation). In contrast,

> for the ODS analysis, performed using the

> classification prior to the Columbia University

> reclassification, the primary outcome was serious

> suicide-related events, comprising events selected

> as possibly suicide-related by each sponsor

> under the search strategy requested by FDA in July

> 2003, and also designated as serious adverse events

> by the sponsors under the standard regulatory

> criteria for " serious. " By focusing on these primary

> outcomes, a comparison of the impact of the two

> systems of case classification is presented.

>

> Comparison of risk estimates obtained with the two

> analyses

>

> Finally, the risk estimates obtained from the two

> analyses are directly compared.

>

> RESULTS

>

> Comparison of statistical methods

>

> Table 1 below displays the relative risks obtained

> by the two analytic methods when identical patient

> populations and classification of cases are used in

> the respective analyses.

>

> Table 1: Comparison of results from two methods for

> sponsor's classification of suicide related events

>

> All sponsor-defined suicide-related events Category

> of trials ODS analysis:

>

> Combined incidence rate ratios*

> DNDP analysis:

> Risk ratios*

> Paroxetine 2.69 (1.20-6.00) 2.47 (1.16-5.27)

> Sertraline 2.03 (0.51-8.16) 1.72 (0.50-5.89)

> Venlafaxine 3.33 (1.08-10.33) 3.03 (1.04-8.80)

> Fluoxetine 0.88 (0.34-2.30) 0.98 (0.38-2.50)

> Citalopram 1.41 (0.66-3.00) 1.49 (0.72-3.06)

> Mirtazapine 0.53 (0.007-41.45) 0.52 (0.003-8.27)

> Nefazodone ? 2.17 (0.23-20.08)

> Fluvoxamine ? 3.31 (0.14-79.67)

> MDD trials 1.81 (1.19-2.77) Not done

> SSRI** MDD trials 1.58 (0.99-2.52) 1.62 (1.03-2.54)

> Non-MDD trials 2.36 (0.67-8.33) 1.93 (0.68-5.45)

> All trials 1.86 (1.25-2.78) 1.81 (1.24-2.64)

> ?Ratio undefined due to zero events in placebo group

> *Mantel-Haenszel method, fixed effects model

> **includes paroxetine, sertraline, fluoxetine,

> citalopram, fluvoxamine

> There is generally good agreement between the two

> methods, suggesting that the findings are not

> sensitive to changes in statistical computing

> methodology.

>

> Comparison of case classifications

>

> The ODS analysis included a total of 78 serious,

> suicide-related events, as defined above. Of these

> 78 cases, 61 (78.2%) were classified by the Columbia

> University group as Outcome 3 (definitive suicidal

> behavior). Of the remaining 17 cases, an additional

> 13 (16.7%) were classified as self-injurious

> behavior with unknown intent (Code 3), and the

> remaining 4 cases were classified in other outcomes.

>

> Conversely, the Columbia University group identified

> a total of 95 cases as definitive suicidal behavior

> (Outcome 3). Of these 95 cases, 61 (64.2%) were

> serious, suicide-related events in the ODS analysis;

> sixteen (16.8%) of the 95 cases were sponsor-defined

> suicide-related but nonserious events, and thus were

> excluded from the ODS primary analysis; and 18 cases

> were new, i.e., were identified through the expanded

> search for cases that was not part of the ODS

> analysis.

>

> On a net basis, the DNDP analyses considered 17 more

> cases than the ODS analysis.

>

> Comparison of risk estimates

>

> Table 2 below compares the risk estimates derived

> from the two analyses, using the abovementioned case

> definitions.

>

> Table 2: Comparison of Columbia University Outcome 3

> with Serious suicide-related events

>

> Category of Trials Total

> N

> Drug

> Total

> N

> Pbo

> Incidence rate

> ratios, serious

> suicide-related

> events

> (ODS analysis)*

> Risk ratios,

> Columbia

> University

> Outcome 3,

> (DNDP analysis)*

>

> Paroxetine 642 549 2.19 (0.92-5.24) 2.65 (1.00-7.02)

> Sertraline 281 279 2.52 (0.49-13.01) 1.48

> (0.42-5.24)

> Venlafaxine 339 342 1.80 (0.52-6.20) 4.97

> (1.09-22.72)

> Fluoxetine 249 209 0.88 (0.32-2.44) 0.92 (0.39-2.19)

> Citalopram 210 197 2.54 (0.91-7.05) 1.37 (0.53-3.50)

> Mirtazapine 170 88 ? 1.58 (0.06-38.37)

> Nefazodone 279 189 ? **

> Fluvoxamine 57 63 ? 5.52 (0.27-112.55)

> Bupropion 71 36 ** **

> All MDD trials 1586 1299 1.95 (1.19-3.21) 1.71

> (1.05-2.77)

> SSRI?? MDD trials 955 843 1.87 (1.10-3.18) 1.41

> (0.84-2.37)

> Non-MDD trials 712 653 1.31 (0.26-6.72) 2.17

> (0.72-6.48)

> All trials 2298 1952 1.89 (1.18-3.04) 1.78

> (1.14-2.77)

>

> *Mantel-Haenszel method, fixed effects model

> **No events in either arm

> ?Ratio undefined due to zero events in placebo group

> ??includes paroxetine, sertraline, fluoxetine,

> citalopram, fluvoxamine

>

> The overall risk estimate for the primary outcome

> for the " all trials " analysis decreased with the

> Columbia University reclassification analysis from

> 1.89 to 1.78; the confidence intervals for both risk

> estimates exclude one. For the category of SSRI MDD

> trials, the risk estimate decreased and lost

> statistical significance with the Columbia

> University reclassification analysis. In terms of

> results for individual drugs, the risk estimates for

> paroxetine and venlafaxine increased.

>

> CONCLUSIONS

>

> Consistent with the analysis described in the

> 3-19-04 consult, the DNDP meta-analysis also

> indicates a statistically significant association of

> suicidal events with antidepressant drug treatment

> in short-term pediatric clinical trials for all

> indications. In terms of subgroups of trials, the

> major differences were that the risk estimate for

> the category of SSRI MDD trials was lower and not

> statistically significant with the DNDP analysis,

> while the risk estimates for two drugs (paroxetine

> and venlafaxine) increased. In all three cases,

> however, the new point estimate falls within the

> confidence limits of the previous result.

>

> RECOMMENDATIONS

>

> With respect to what further analyses might be

> undertaken with the Columbia University dataset, I

> propose an analysis that examines events occurring

> after treatment discontinuation, since there appears

> to be a signal for at least paroxetine in this

> regard (please refer to the previous consults by the

> undersigned for details).

>

> Beyond what else may be done with the current

> dataset, I also propose an analysis with psychiatric

> inpatient hospitalization as the outcome. While not

> specific for suicidal behavior, this might give

> insight into more general behavioral toxicities, and

> would have the advantage of being easily determined

> from the existing case reports. In addition, I agree

> with the plans to analyze the new data from the NIMH

> Treatment of Adolescent Depression Study (TADS),

> which will provide additional data for fluoxetine.

> With respect to possible regulatory actions, please

> refer to my recommendations in the 3-19-04 consult;

> the results from the DNDP meta-analysis using the

> Columbia University reclassification do not

> materially affect the recommendations I made

> previously.

>

> (hard copy signed 8-16-04)

> Andrew D. Mosholder, M.D., M.P.H.

> Epidemiologist

> (hard copy signed 8-16-04)

> Mary Willy, Ph.D.

> Epidemiology Team Leader

>

> [Non-text portions of this message have been

> removed]