`
`CENTER FOR DRUG EVALUATION AND
`RESEARCH
`
`
`APPLICATION NUMBER:
`22-272Orig1s014
`
`STATISTICAL REVIEW(S)
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`U.S. Department of Health and Human Services
`Food and Drug Administration
`Center for Drug Evaluation and Research
`Office of Translational Sciences
`Office of Biostatistics
`
`
`S T A T I S T I C A L R E V I E W A N D E VA L U A T I O N
`POST-MARKETING OBSERVATIONAL STUDIES
`
`22-272
`
`NDA #:
`
`Drug Name:
`OxyContin (oxycodone hydrochloride)
`Indication(s):
`Management of moderate to severe pain
`Applicant:
`Purdue Pharma L.P.
`Review Date(s):
`October 5, 2012
`Review Priority:
`Standard
`Biometrics Division:
`Division of Biometric 7
`Statistical Reviewer:
`Eric Frimpong, PhD
`Concurring Reviewers: Yu-te Wu, PhD
`Mark Levenson, PhD, Deputy Director
`Division of Anesthesia, Analgesia, and Addiction Products (DAAAP)
`Bob Rappaport, M.D., Director
`Sharon Hertz, M.D., Deputy Director
`Ellen Fields, M.D., Team Leader
`Pamela Horn, M.D., Team Leader
`Jin Chen, M.D., PhD, Medical Officer
`Epidemiology Division: Division of Epidemiology II, Office of Surveillance and Epidemiology
`(OSE)
`Maloney, Elizabeth, MS, DrPH, Team Leader
`Trinidad James MPH, MS, Epidemiologist
`Cyndy Kornegay, Ph.D. Epidemiologist
`Cathy Dormitzer, Ph.D. Epidemiologist
`Project Manager:
` Lisa Basham, MS
`Keywords: Observational studies, Drug abuse rates, GEE, interrupted time series
`
`
`
`Medical Division:
`Clinical Team:
`
`Epidemiology Team:
`
`Reference ID: 3215400
`
`
`
`3
`
`
`Table of Contents
`1 EXECUTIVE SUMMARY .................................................................................................................................3
`2
`INTRODUCTION ...............................................................................................................................................5
`2.1 BACKGROUND .....................................................................................................................................................5
`2.2 MATERIAL REVIEWED...........................................................................................................................................6
`2.3 STUDY OVERVIEW................................................................................................................................................6
`STATISTICAL EVALUATION ........................................................................................................................7
`3.1 STUDY 1 (NAVIPPRO STUDY)...........................................................................................................................7
`3.1.1
`Study Overview ......................................................................................................................................7
`3.1.2
`Study Design and Outcome Measures ...................................................................................................7
`3.1.3
`Statistical Methodologies.......................................................................................................................8
`3.1.4
`Results and Conclusions ........................................................................................................................9
`3.1.5
`SUMMARY AND CONCLUSIONS........................................................................................................9
`3.2 STUDY 2 (KAISER STUDY)...............................................................................................................................10
`3.2.1
`Study Overview ....................................................................................................................................10
`3.2.2
`Study Design and Outcome Measures .................................................................................................10
`3.2.3
`Statistical Methodologies.....................................................................................................................11
`3.2.4
`Results and Conclusions ......................................................................................................................11
` 3.2.4 SUMMARY AND CONCLUSIONS .......................................................................................................12
`3.3 STUDY 6 (PMPS STUDY) ..................................................................................................................................12
`3.3.1
`Study Overview ....................................................................................................................................12
`3.3.2
`Study Design and Outcome Measures .................................................................................................12
`3.3.3
`Statistical Methodologies.....................................................................................................................13
`3.3.4
`Results and Conclusions ......................................................................................................................13
`3.3.5
`SUMMARY AND CONCLUSIONS......................................................................................................14
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Reference ID: 3215400
`
`2
`
`
`
`
`1 EXECUTIVE SUMMARY
`
`OxyContin (OC) was first approved by the Agency on December 12, 1995. OxyContin is a
`schedule II controlled substance with label indication, “For the management of moderate to
`severe pain where use of an opioid analgesic is appropriate for more than a few days.”
`Oxycodone products are common targets for both drug abusers and drug addicts. The Agency
`approved the new reformulated OxyContin (ORF) in April 5, 2010. The new formulation of
`OxyContin was designed to make breaking, dissolving, crushing or chewing the tablet more
`difficult. Purdue ceased shipping the original formulation of OxyContin on August 5, 2010 and
`began shipping only reformulated tablets from August 9, 2010. As of January 2011, more than
` of filled prescriptions for OxyContin were reformulated OxyContin. At the Joint Meeting of
`the Anesthetic and Life Support Drugs Advisory Committee and the Drug Safety and Risk
`Management Advisory Committee on October 21 and 22, 2010, Purdue proposed multiple post-
`marketing studies to assess the effects of reformulated OxyContin in the setting that reflects the
`actual usage. Purdue submitted preliminary reports on the studies as May 2012 to the Agency on
`July 31, 2012.
`
`The Division of Epidemiology II requested the Division of Biometrics VII to review the
`preliminary report submitted in July 2012. This review provides a statistical evaluation of the
`design, methods and proposed analyses for studies 1, 2 and 6. An assessment of the preliminary
`results is also provided. However, a thorough and complete evaluation of the study results should
`be conducted upon the completion of the studies. A separate biostatistical review by Dr. Zhang
`addresses studies 3, 4, 5, and 11.
`
`Study 1 was designed to investigate the routes and rates of OxyContin abuse among patients in
`substance abuse treatment programs in the ASI-MV Connect NAVIPPRO System. Specifically,
`patterns of past 30-day abuse of reformulated OxyContin (ORF) are compared to those of the
`original formulation (OC) after the introduction of ORF. In addition, the study assessed abuse
`through routes of administration (ROA) that require tampering, particularly snorting, injecting,
`and smoking. These were compared to original OxyContin and comparator opioids. The report
`covers preliminary analyses of the data from June, 1, 2009 to March, 31, 2012.
`
`Generalized linear mixed model was used to estimate pre-ORF and post-ORF period percentages
`and relative percent change from the pre to post ORF period. Specifically, quarterly prevalence
`of past 30-day abuse for OC and ORF were compared to changes in comparator opioid
`analgesics ER morphine and ER oxymorphone. Although the data presented for 6 quarters post
`ORF is consistent with the study hypotheses of lower rates of abuse ORF, its profile compared to
`OC beyond the second quarter is very similar. The preliminary results show a considerable drop
`in levels of abuse through both oral and non-oral (smoking , snorting and injecting) after the
`introduction of ORF; however, with limited data points, the results do not support any substantial
`long term pattern.
`
`Study 2 investigated the changes in rates of opioid overdose and poisoning (OOP) among
`patients dispensed OxyContin or comparator opioids in the Kaiser Permanente Northwest and
`Northern California regional health care systems before and after the introduction of
`
`
`3
`
`Reference ID: 3215400
`
`(b) (4)
`
`
`
`
`reformulated OxyContin (ORF). The study used chart abstraction data from February 2003 to
`July 2010, and 15 months following the introduction of ORF. The rates of OOP event associated
`with OxyContin use were compared to three groups of comparator opioids: a) other extended
`release opioids, b) immediate release, single entity oxycodone, and c) all other prescription
`opioids.
`
`The findings at the time of this report do not suggest any substantial changes in dispense patterns
`or abuse rates or both. Data were only available for one full six-month period following the
`transition from the OC to ORF at this point. Limited data in the post-ORF period precludes the
`adequate assessment of the study results.
`
`Study 6 used data from the Ohio Prescription Monitoring Program (PMP) and IMS LRx
`prescription database. The PMP study examined the number of individuals who obtained
`prescriptions by multiple prescribers and filled at multiple pharmacies. In the IMS LRx analysis,
`the goal was to assess the potential changes in the proportion of opioid shopping behavior among
`OxyContin users after the introduction of ORF. Doctor shopping was defined as a patient that
`visits multiple prescribers and pharmacies to obtain and fill more than necessary opioid
`prescriptions, in order to abuse or sell the excess opioids.
`
`The PMP analysis consisted of data from August 8, 2008 to June 11, 2011. The IMS LRx
`analysis consisted of 2 six-month pre-periods (July to December 2009 and January to June 2010)
`and 2 six-month post-periods (January to June 2011 and July to December 2011). The PMP
`analysis used data from the Ohio Automated Rx Reporting System (OARRS). The study
`estimated the counts and rates of individuals who filled OxyContin prescriptions from a
`combination of 1-5 or more prescribers and 1-5 or more pharmacies. In IMS LRx analysis, the
`study used a database that consisted of patient de-identified longitudinal prescription from a
`sample of IMS Health retail and mail order prescriptions universe. Relative change in
`proportions was used to assess the shopping behavior of OxyContin from pre-ORF to post-ORF.
`
`There are a total of 5 data points in the PMP analysis and 4 in the IMS LRx analysis. With very
`few data points, the analysis does not provide sufficient information to identify or establish a
`trend.
`
`The design aspects of post-marketing observational studies on abuse deterrence were discussed
`in the Joint Meeting of the Anesthetic and Life Support Advisory Committee and the Drug
`Safety and Risk Management Advisory Committee in October 2010. The trend approach and the
`requirement of a sufficient period of time to establish the pattern of abuse and to demonstrate
`sustainability were emphasized by the committee. In order to properly characterize the abuse
`pattern over time, we need to be confident that the trend is stable and well characterized, which
`may require longer observation periods and the ability to consider the autocorrelation structure
`and possibly periodicity or seasonal patterns in the data. The accuracy, in terms of bias and
`variability, of the outcome measure would also affect the necessary length of the observational
`period. The three studies covered in this review had approximately 1 to 1.5 years of data after
`ORF was introduced into the market, corresponding to 2 to 6 data points depending on the data
`source. The adequacy of data points/structure for these studies should be further evaluated upon
`
`
`4
`
`Reference ID: 3215400
`
`
`
`
`the completion of the study. Therefore, the results presented in this preliminary study report do
`not provide conclusive evidence for the evaluation of abuse deterrence.
`
`
` INTRODUCTION
`
`2.1 Background
`
` 2
`
`
`
`
`OxyContin (OC) was first approved by the Agency on December 12, 1995. OxyContin is a
`schedule II controlled substance with label indication, “For the management of moderate to
`severe pain where use of an opioid analgesic is appropriate for more than a few days.”
`Oxycodone products are common targets for both drug abusers and drug addicts. The Agency
`approved the new reformulated OxyContin (ORF) in April 5, 2010. The new formulation of
`OxyContin was designed to make breaking, dissolving, crushing or chewing the tablet more
`difficult. Purdue ceased shipping the original formulation of OxyContin on August 5, 2010 and
`began shipping only reformulated tablets from August 9, 2010. As of January 2011, more than
` of filled prescriptions for OxyContin were reformulated OxyContin. At the Joint Meeting of
`the Anesthetic and Life Support Drugs Advisory Committee and the Drug Safety and Risk
`Management Advisory Committee on October 21 and 22, 2010, Purdue proposed multiple post-
`marketing studies to assess the effects of reformulated OxyContin in the setting that reflects the
`actual usage. Purdue submitted preliminary reports on the studies as May 2012 to the Agency on
`July 31, 2012.
`
`As part of the post marketing requirement, the sponsor has conducted six epidemiology studies
`to assess the effects of ORF on patterns of abuse and misuse, and their consequences of
`addiction, overdose, and death. Additionally, the sponsor has conducted five supplemental
`studies or analyses of surveillance systems that provide additional information on the effects of
`ORF. The epidemiology studies were designed to assess the effects of ORF on patterns of abuse
`and misuse, and their consequences of addiction, overdose and death. Purdue submitted reports
`on the studies as May 2012 to the Agency on July 31, 2012.
`
`The Division of Epidemiology II requested the Division of Biometrics VII to review the
`preliminary report submitted in July 2012. The purpose of this statistical review is to provide
`comments on the statistical approaches and results for the three observational studies as below:
`
`
`• Study 1: Routes and Rates of OxyContin Abuse Among Patients in Substance Abuse
`Treatment Programs in the ASI-MV Connect NAVIPPRO System
`• Study 2: Changes in Rates of Opioid Overdose and Poisoning Events in the Kaiser
`Permanente Health System with the Introduction of Reformulated OxyContin
`• Study 6: Doctor-shopping for OxyContin as Measured by Prescription Monitoring
`Programs
`
` separate biostatistical review by Dr. Zhang addresses studies 3, 4, 5, and 11.
`
`5
`
` A
`
`
`
`
`
`Reference ID: 3215400
`
`(b) (4)
`
`
`
`2.2 Material reviewed
`
`
`OxyContin (oxycodone hydrochloride controlled-release) Tablets - Report on the Findings as of
`May 2012: Post-marketing Epidemiology Study Program to Assess the Effects of Reformulated
`Oxycontin on Patterns of Abuse and Misuse and their Consequences (Addiction, Overdose and
`Death), Patient Adverse Events, and Unintentional Exposures, July 2012, submitted on July 31,
`2012 for NDA #022272.
`
`
`
`
`
`
`
`
`2.3 Study Overview
`
` Study 1:
`This study was designed to investigate the routes and rates of OxyContin abuse among
`patients in substance abuse treatment programs in the ASI-MV Connect NAVIPPRO
`System. Specifically, patterns of past 30-day abuse of reformulated OxyContin (ORF)
`are compared to those of the original formulation (OC) after the introduction of ORF. In
`addition, the frequency of use of ORF as measured by number of days per month used
`was compared to that observed for original OxyContin and comparator opioids. In
`addition, the study assessed abused through routes of administration (ROA) that require
`tampering, particularly snorting, injecting, and smoking. These were compared to original
`OxyContin and comparator opioids. The report covers a preliminary analysis of the data
`from June, 1, 2009 to March, 31, 2012.
`
` •
`
`• Study 2:
`The study investigated the changes in rates of opioid overdose and poisoning (OOP) in
`the Kaiser Permanente Health System before and after the introduction of reformulated
`OxyContin (ORF). The study used chart abstraction data from February 2003 to July
`2010 and 15 months following the introduction of ORF. Overall there were
` events,
` Kaiser Permanente Northwest (KPNW) and
` from Kaiser Permanente Northern
`California (KPNC).
`
`The objective was to estimate and compare rates of opioid overdose and poisoning (OOP)
`events before and after the introduction of ORF among individuals dispensed OxyContin.
`The rates are compared to individuals dispensed to three groups of comparator opioids: a)
`other extended release opioids, b) Immediate release, single entity oxycodone, and c) all
`other prescription opioids.
`
`• Study 6:
`This study used data from the Ohio Prescription Monitoring Program (PMP) and IMS
`LRx prescription database. The PMP study examined the number of individuals who
`obtained prescriptions by multiple prescribers and filled at multiple pharmacies. In the
`IMS LRx analysis, the study used a database that consisted of patient de-identified
`longitudinal prescription from a sample of IMS Health retail and mail order prescriptions
`universe. The goal was to assess the potential changes in the proportion of opioid
`shopping behavior among OxyContin users after the introduction of ORF. Doctor
`shopping was defined as a patient that visits multiple prescribers and pharmacies to
`
`6
`
`Reference ID: 3215400
`
`(b) (4)
`
`(b) (4)
`
`(b) (4)
`
`
`
`
`
`obtain and fill more than necessary opioid prescriptions, in order to abuse or sell the
`excess opioids.
`
`The PMP analysis consisted of data from August 8, 2008 to June 11, 2011 (5 data points).
`The IMS LRx analysis consisted of 2 six-month data in the pre ORF period (July to
`December 2009 and January to June 2010) and 2 six-month data in the post-ORF period
`(January to June 2011 and July to December 2011).
`
`
`
` 3
`
`
`3.1
`
` STATISTICAL EVALUATION
`
`Study 1 (NAVIPPRO STUDY)
`
`Study Overview
`
`3.1.1
`
`This study was designed to investigate the routes and rates of OxyContin abuse among patients
`in substance abuse treatment programs in the ASI-MV Connect NAVIPPRO System.
`Specifically, patterns of past 30-day abuse of reformulated OxyContin (ORF) are compared to
`those of the original formulation (OC) after the introduction of ORF. In addition, the frequency
`of use of ORF as measured by number of days per month used was compared to that observed
`for original OxyContin and comparator opioids. In addition, the study assessed abused through
`routes of administration (ROA) that require tampering, particularly snorting, injecting, and
`smoking. These were compared to original OxyContin and comparator opioids. The report
`covers preliminary analyses of the data from June, 1, 2009 to March, 31, 2012.
`
`3.1.2 Study Design and Outcome Measures
`
`
`This was an observational study designed to compare the prevalence, prescription-adjusted
`prevalence rates and route of administration (ROA) patterns of past 30-day abuse of ORF to that
`of OC before and after the introduction of ORF. These estimates were compared to changes in
`comparator opioid analgesics ER morphine and ER oxymorphone in the same period. The study
`used a stratified two-stage cluster design to sample respondents. First, the number of sites was
`determined. Second, the number of patients within sites that are needed to ensure a representative
`sample was obtained. Each respondent reported the abused compound and the route(s) in which
`the compound was abused. Past prevalence of abuse and ROA of OC measured from June 1,
`2009 through August 8, 2010, about 5 quarterly data points were compared with ORF experience
`from August 9, 2010 through March 31, 2012 (6 quarterly data points).
`
`The study examined the ROA patterns and abuse rates of ORF by four outcome measures:
`• prevalence of past 30-day abuse among all respondents evaluated or within the subset of
`individuals reporting past 30-day abuse of any prescription opioid
`• prescription-adjusted prevalence rates of abuse
`• prevalence of abuse via oral and non-oral ROA for ORF, OC and comparator opioids
`•
`frequency of abuse
`
`
`
`Reference ID: 3215400
`
`7
`
`
`
`
`
`Abuse and ROA patterns were captured via self-report during the ASI-MV interview which
`contains product-specific questions about abuse, routes and sources.
`
`Comments:
`The study employed an observational design that gathered information on opioid abusers before
`and after the introduction of the reformulated OxyContin. The proposed design is appropriate if
`limitations such as misclassification of abused compound; and selection bias due to the sample
`of sites are minimal. Although, we can capture the information of specific products and routes by
`self-report, social desirability bias is a problem with self-report measures and can affect the
`validity of the study.
`
`3.1.3 Statistical Methodologies
`The study used generalized linear mixed (GLMM) models to estimate pre-ORF and post-ORF
`period percentages and relative percent change from the pre to post ORF period. Logistic
`regression models, using the GLIMMIX procedure, were used to evaluate the percentages and
`relative percent change from the pre to post ORF. Also, log-binomial regression model was used
`to estimate the mean number of days of abuse and relative percent change in the mean number of
`days of abuse for pre-ORF and post-ORF period. The independent variables for the regression
`models include the main effects, two and three way interactions terms of opioid/drug indicator,
`ROA indicator, and time (per quarter). Random effects were used to account for multiple
`observations per ASI-MV respondent and nesting of respondents within a zip code.
`
` Comments:
`1. Since the study collected repeated observations on respondents over time and clustered
`observations within sites, the GLMMs are appropriate to estimate the population-averaged
`outcome. More specifically, the outcome was the average change in respondents’ responses
`before and after the introduction of ORF.
`
`2. The form of the dependency (within respondents) does not usually affect parameter estimates
`as long as the regression models are correctly specified, however, we still recommend the
`sponsor to conduct sensitivity analyses for different specifications of the intra-cluster correlation
`matrix to assess the robustness of the study outcomes.
`
`
`3. The analyses of seasonal effects were neither discussed in the protocol nor the interim report.
`GLMM model can handle seasonal effect either through various covariance structures or using
`sine cosine pairs in the model. The sponsor should first clarify if seasonal or temporal pattern
`exists in the data, if it does exist, then the appropriately scaled harmonic functions should be
`considered.
`
`4. The dependent variables include time (per quarter). There are no discussions on how the time
`variable enters the model. Sponsor should clarify this and should also consider the
`transformations of time variable to properly capture the trends using the GLIMMIX procedure.
`
`
`
`Reference ID: 3215400
`
`8
`
`
`
`
`
`3.1.4 Results and Conclusions
`
`
`As indicated in the study report, the trend of OxyContin abuse (OC and ORF) via any ROA over
`the study period declined in the quarterly prevalence of past 30-day abuse following the
`introduction of ORF as a proportion of all assessments (Figure 9, page 34) and among
`prescription opioid abusers (Figure 10, page 35). With respect to ROA, the analysis yielded
`similar findings (Table 9, page 35; and Figures 11, 12, and 13, pages 39-41).
`
`Comments:
`We should interpret sponsor’s results with caution for reasons stated below:
`
`1. Although, the levels of abuse declined in the first three quarters of post-ORF period, the levels
`of OC and ORF thereafter remains almost the same and showed consistent patterns for both OC
`and ORF. The abuse rate of OC in the post-ORF period is expected to drop gradually over time
`because of the limited supply. As stated in the report, prescriptions filled at pharmacies for
`original OxyContin constituted
` of total OxyContin prescriptions in
`January 2011, June 2011 and December 2011, respectively. The limited supply of original
`OxyContin from prescriptions filled at pharmacies is unlikely to account for the continued levels
`of abuse of OC. Some degree of misclassification between OC and ORF exists; therefore, the
`study results are subject to misclassification bias.
`
`2. The study results are based on an interim analysis and should be interpreted within the time
`frame and limited data provided. An explicit explanation on how the sponsor intends to further
`investigate this issue is encouraged. As noted in the study report, the results are preliminary and
`abuse patterns may change over time. Therefore, a complete assessment of pre and post ORF
`may require a longer period.
`
`3.1.5 SUMMARY AND CONCLUSIONS
`Self-report captures specific products and routes but social desirability bias is a problem with
`self-report measures and can affect the validity of the study. Although the use of ASI-MV
`sentinel surveillance sample provides a sensitive population with a high potential of drug abuse,
`it is not a random sample. Therefore, the preliminary results and results upon the completion of
`the study may not be generalized to broader population.
`
`Although the data presented for 20 months post ORF is consistent with the study hypotheses of
`lower rates of abuse ORF, its profile compared to OC beyond the second quarter is very similar.
`The preliminary results shows a considerable drop in the level of ROA after the introduction of
`ORF, however the results do not support any substantial long term pattern. Because the
`observation period is short, long term patterns can not be assessed with the preliminary data.
`
`The statistical analysis employed in the study appears appropriate for the study design. However,
`it is not clear how the time variable is used to capture trends in the data. Also, the analyses on
`seasonal effects with respect to the GLMM model are not discussed. Since data were collected
`from
` centers with multiple reports of routes of administration from respondents, sensitivity
`analysis of different specifications of the intra-cluster correlation matrix are recommended.
`
`
`9
`
`Reference ID: 3215400
`
`(b) (4)
`
`(b) (4)
`
`
`
`
`
`In addition to the results presented in the Table 10, it is recommended that the investigators
`include the number of unique respondents by quarter. Also, investigators may add summaries of
`unique respondents that contributed to multiple ROA.
`
`3.2
`
`Study 2 (KAISER STUDY)
`
`3.2.1 Study Overview
`The study investigated the changes in rates of opioid overdose and poisoning (OOP) in the
`Kaiser Permanente Health System before and after the introduction of reformulated OxyContin
`(ORF). The study used chart abstraction data from February 2003 to July 2010 and 15 months
`following the introduction of ORF. Overall there were
` events,
` Kaiser Permanente
`Northwest (KPNW) and
` from Kaiser Permanente Northern California (KPNC).
`
`The objective was to estimate and compare rates of opioid overdose and poisoning (OOP) events
`before and after the introduction of ORF among individuals dispensed OxyContin. The rates for
`individuals dispensed to OC/ORF are compared to those for individuals dispensed to three
`groups of comparator opioids: a) other extended release opioids, b) Immediate release, single
`entity oxycodone, and c) all other prescription opioids.
`
`
`3.2.2 Study Design and Outcome Measures
`An interrupted time series design was used to longitudinally compare rates of OOP events
`associated with OxyContin to rates of OOP events associated with other oxycodone and opioid
`formulations over a 10-year period. Specifically, an interrupted time series approach and a ratio
`of risk ratios approach is used to compare trends in rates of OOP events before and after the
`introduction of ORF. For the proposed ITS analysis, abuse rates of OxyContin in six-month
`interval would be compared to immediate-release single ingredient oxycodone, other long-acting
`opioids, and other Schedule II opioids. The computed rates cover a period of seven years before
`ORF and 2.5 years post-ORF.
`
`The poisonings and overdoses rates associated with OxyContin were computed as follows:
`• Number of poisonings/overdoses for people with a dispense of OxyContin/oxycodone ER
`divided by the number of people with a dispensing of OxyContin/oxycodone ER and
`• Number of poisonings/overdoses for people with a dispense of OxyContin/oxycodone ER
`divided by Morphine equivalent milligrams of all dispenses of OxyContin/oxycodone ER
`A proposed analysis to examine the rate of OOP events per person time exposed to
`OxyContin is defined as
`• The ratio of the number of poisonings/overdoses for people with a dispensing of
`OxyContin and person time on OxyContin
`
`
`Person time on OxyContin is defined as the number of days on OxyContin for people dispensed
`OxyContin in a given six-month period. Similar rates were computed for immediate release
`single ingredient oxycodone, other class REMS opioids, and other Schedule II opioids for each
`
`
`
`Reference ID: 3215400
`
`10
`
`(b) (4)
`
`(b) (4)
`
`(b) (4)
`
`
`
`
`six-month period. The three specified rates will be used to determine the changes in trends of the
`rate of OOP events after the new formulation.
`
`Comments:
`
`1. The proposed study design compares the rates of OOP events in OxyContin to that of other
`opioids after the introduction of ORF. The comparable opioids were neither reformulated before
`or after the introduction of ORF. Therefore, the design appropriately separates the effects due to
`the introduction of ORF from the effects that may have occurred at that time.
`
`2. An important measure of the interrupted time series analysis is the difference between the
`predicted behavior in post- ORF phase (using pre-ORF series) and the actual (observed)
`behavior of the series in the post- ORF phase. Therefore, we recommend that the sponsor
`present the results (predicted estimates) of post-ORF using pre-ORF data and summaries of the
`actual post-ORF data.
`
`
`3.2.3 Statistical Methodologies
`The study proposes to use an interrupted time series (ITS) approach and a ratio of risk ratios
`approach (RR). The ITS approach models the rate of OOP from February, 2003 to July, 2010 as
`phase 1 and compares its estimates to that of phase 2 from August, 2010 to December, 2012.
`
`Comments:
`
`In the proposed ITS approach, rates are calculated in six-month intervals for a period of 10
`years. Seven years of data prior to ORF introduction are compared to 2.5 years after ORF, i.e.,
`14 data points in pre-ORF vs. 5 data points in post-ORF. In order to properly characterize the
`abuse pattern over time, sufficient number of data point is required to account sufficiently
`characterized the trend and serial dependency (autocorrelation), and also possibly the seasonal
`or temporal pattern in the data. The accuracy, in terms of bias and variability, of the outcome
`measure would also affect the observational period. Simulation studies (Crosbie, 1993) indicates
`that the estimate of autocorrelation is unreliable with fewer data points, leading to an inflated
`type II error, i.e., insufficient power. The adequacy of data points/structure for this study will be
`further evaluated upon the completion of the study.
`
`3.2.4 Results and Conclusions
`
`
`According to the study report, “ At the time of this report, data were only available for only one
`full six- month period following the transition from the original to the new formulation of
`OxyContin.” Prior to the transition period, the dispense patterns of OxyContin showed an
`upward trend from 2003 to 2008. However, the pattern declined rapidly through the transition
`period and thereafter (Figures 14 and 18, pages 64 and 69 respectively). Therefore, the only data
`point post-ORF may not be