(PROMPT): Self-Controlled Design, Technical Users - Mini
Contents
1. Table 12 INCEVENT N 102 103 103 LA R The other kind of HOI exclusion that can be specified is if some other health outcome occurred within a certain period of time prior to the one of interest To specify this one goes to the EVENT CODES table and sets EVENT TBL FLG to 0 and INCEVENT TBL FLAG to 1 In our example studies we did not have Mini Sentinel PROMPT Technical Users Guide 11 Self Controlled Design Tool Documentation M CO HOI exclusions based on other health outcomes If we had we would have added them as SUBGROUPs to the EVENT CODES and INCEVENT tables Finally the exposure period for which data are to be extracted from the M S distributed dataset must be specified for each data partner in the EXPOSURE PERIODS table Specifying exposure date range does not restrict the follow up period only the dispensing or exposure period Because of differences among data partners with respect to date ranges frequency of data updates and completeness of data in each update etc it will likely be appropriate to select different exposure date ranges for different Data Partners Also it might not be possible to pre specify all date ranges within each Data Partner into the future For each new data update from a specific Data Partner a date range immediately following the date range used the last time that Partner s data were extracted and analyzed should be chosen It is important that there be no overlap i2. LISINOPRIL CLINDAMYCIN MMRV 1 MMRV 2 NOTES Exposure ACEi s ARBs Antibiotics Specify if exclusions direct renin patients with inhibitors certain exposures on the same day as or in the x days prior to the exposure of interest are to be excluded Exposure look 0 183 days 0 183 days 0 183 days back period from dispensing date Health outcome Angioedema AMI Seizures HOI HOI settings All Inpatient Inpatient ED Inpatient ED Possible settings inpatient has inpatient has in which to look higher priority higher priority for potential within a day within a day cases include inpatient ED and clinic outpatient HOI exclusions Prior Prior AMI Prior seizures angioedema HOI exclusion All IP All For MMRV settings cases of interest are those in inpatient and ED settings but only if there were no cases in the prior 183 days in any setting including clinic Mini Sentinel PROMPT Technical Users Guide 3 Self Controlled Design Tool Documentation M iud LISINOPRIL CLINDAMYCIN MMRV 1 MMRV 2 NOTES HOI look back 1 183 days 1 28 days 1 183 days The case must period from be the patient s diagnosis date first one to occur in this period of time in order to count Risk window 1 28 days 1 14 days 7 10 days This is the post exposure Day O period during which an increased risk of the HOI is considered possible and will be examined Comparison 35 to 8 15 to 28 1 to
3. returned by the Data Partners in the extraction table Events that occur in the risk window will be called cases and events that occur in the control window will be called controls An example of output returned from a Data Partner is below Mini Sentinel PROMPT Technical Users Guide 31 Self Controlled Design Tool Documentation E ciue Figure 10 SAS Universal iewer E File Tools Window Help Address Library Hl Freeze M Hide Di show DS Format E Fiter A Font Find NM Events in Events in DP Exposure Start Date End Date Total Events Control Window Risk Window ew Exposure Period Exposure Period Controls Cases O1SEP2012 1 LISINOPRIL 30SEP2012 ze 1 5 2 CUNDAMYON mseP2m2 30sEP2D12 1 2 1 o Table 3 MMRV DISEP2012 305EP2012 eh 0 _4 MMRY 2 DISEP2012 30SEP2012 H ol 0 The analysis will be executed by entering values into the function Analyze Binomial in the R Console window with a case sensitive template as follows Analyze Binomial namez test cases lt controls lt where name name of the text file which will store all associated cumulative tests associated with the unique parameters for each Drug HOI pair test number of the test whether it is the 1 27 3 etc cases lt numeric value from extraction table of the HOI events that occur in the risk window when entering the number leave one
4. DAYS FROM EXPOS TO EVENT MAN DAYS FROM EXPOS TO EVENT MIN DAYS ENR AFTER EXPOS MAX DAYS ENR AFTER EXPOS SUM EXPS PERIOD FIN Summary datasets for adverse events on exposure period summarization level without gender and age stratification This is the dataset that will be used as input by R program Summarization level DPID SITEID GROUP EXPOSURE PERIOD CD EXPOS PERIOD STRT DT Statistics CNT EXP EPISODES CNT IN COMP WND CNT IN RISK WND Mini Sentinel PROMPT Technical Users Guide 20 Self Controlled Design Tool Documentation M iud 2 Program Steps e Import mapping and lookup tables supplied with the package e Dosomeinitial processing of mapping and lookup tables e These include setting some defaults joining some of the tables together and creating e macro variables for min and max dates which will be used latter in the program e Create tables with complete set of all NDC PX and DX codes ever used by the MSCDM data warehouse at a given site e This will be used latter to create an explicit list of codes of interest without wild cards and e inthe format used by a site for example with or without a period in DX or PX codes e Extract medical claims from the diagnosis and procedure files e Extract drug claims from the outpatient pharmacy file e Combine NDC DX and PX data extracts e Summarize RX data for dispenses on the same date e Keep one record per combination of MSCDM variables that suppose to be uni
5. MSCDM tables are in the relational database and you have permission to create temporary tables in the same database If this is the case set TMP2 to the library associated with a schema in the database in which you can create tables This can significantly improve performance ENEE EE LKP_FOR_ALL_CODES_FILE Macro parameter Lookup table that has complete set of code for columns used by various mapping tables below Mini Sentinel PROMPT Technical Users Guide 14 Self Controlled Design Tool Documentation M iud COLUMNS COLUMN NAME Name of the code column used in mapping tables defined below For example GROUP QUERY GRP etc CODE VALUE Value for numeric code CODE VALUE C Value for character code CODE DESC Code description EE GROUPS MAPPING Macro parameter Mapping table between study group code GROUP and QUERY GRP and EVENT GRP COLUMNS GROUP Code for a study group A code which identifies both population selection QUERY and adverse effects EVENT set of conditions Numeric 1 2 3 Required QUERY GRP A code which identifies set of conditions for population selection Usually defines population who had exposures to a set of drugs Numeric 1 2 3 Required EVENT_GRP A code which identifies set of conditions for adverse events of interest Numeric 1 2 3 Required EXPOSURE_PERIODS_FILE Macro parameter Mapping table for exposure periods The dataset is distributed as
6. Private section Private DPID SitelD 3 notify the MSOC 6 For KP Mini Sentinel Programmers Only Instead of uploading the results to the Mini Sentinel secure portal please place the SAS datasets and log files in a zip file and post them on to the Mini Sentinel Content Area on the KP Secure File Transfer Web Site https www kpchr org securefiletransfer apps default aspx 7 Workplan Timeline Please complete by June 25 2013 8 Package Documents 1 main SAS Program sasprograms subdirectory mini sentinel to08 scd wp1 b2 sas 2 supporting SAS Programs files inputfiles subdirectory mp4 10 sas ms episoderec sas 10 lookup files inputfiles subdirectory exposure periods sas7bdat event sas7bdat incevent sas7bdat incquery sas7bdat Ikp for all codes sas7bdat age groups sas7bdat groups mapping sas7bdat event codes sas7bdat query sas7bdat query codes sas7bdat This workplan document and technical specification file mini sentinel toO8 scd wp1 b2 workplan pdf Timeframe for Data to be Included April 1 2013 June 30 2013 MSCDM Files Accessed Demographic Diagnosis Dispensing Mini Sentinel PROMPT Technical Users Guide 46 Self Controlled Design Tool Documentation egen CIE Procedure EnrollmentOutput Files This program generates 1 3 SAS datasets 2 1 log file and 3 1 Ist file to the msoc subdirectory Please return all 3 files to MSOC 3 SAS output datasets returned to MSOC amp dpid amp site
7. a part of the package In addition to the column EXPOSURE_PERIOD_CD the dataset has two columns EXPOS PERIOD STRT DT and EXPOS PERIOD END DT These dates can be different for different data partner sites depending on sites data availability and refresh rate For example the exposure period can be a month length for one site and a quarter for another This is possible because the study compares two statistics calculated at same data partner site the number of adverse outcomes in risk time window with the number of adverse outcomes in comparison time window The time periods appropriate to the site should be supplied by MINI SENTINEL OPERATIONS CENTER or updated by local site per leading site instructions COLUMNS EXPOSURE PERIOD CD Numeric 1 2 3 EXPOS PERIOD END DT SAS date Mini Sentinel PROMPT Technical Users Guide 15 Self Controlled Design Tool Documentation M iu EXPOS PERIOD STRT DT SAS date QUERYFILE Parameter Mapping table for population selection variables QUERY COLUMNS QUERY GRP A code which identifies set of conditions for population selection Usually defines population who had exposures to a set of drugs Numeric 1 2 3 Required SUBGROUP Identifies group of codes typically a set of drugs Numeric 1 2 3 Required WASHTYP Selects how incidence events will be defined Character 3 Possible values the same as in MP3 MIN SING MULT WASHPER Length of event washo
8. amp Espected E Cases o gt cL w pont o 8 L DO J N 5 d 3 e ES 1 2 Test Test Observed Relative Risk Log likelihood ratio 1 D e o m ue gt Soc 2 o O qm 9 e z E o c oN ZS a S a pr oO 1 2 1 2 Test Test Mini Sentinel PROMPT Technical Users Guide 38 Self Controlled Design Tool Documentation egen Wee 3 Analyze Binomial name TestFile test 3 cases lt 7 controls lt 1 Test 3 is explained in figures 17 and 18 Figure 17 File Edit View Misc Packages Windows Help lea ale gt Analyze Binomial name TestFile test 3 cases lt 7 controls 1 gt Reject HO No further sequential analyses are needed 90 Confidence interval for the actual relative risk 1 2 10 4 Cumulative alpha spent Test Cases Controls Cases Controls E Cases RR LLR target actual CV HO rejected i 5 1 5 1 3 51 456 0 0156 0 0156 6 No 2 3 0 8 1 4 5 63 099 0 0156 0 0156 9 No 3 7 L 15 2 8 5 7 5 5 626 0 0259 0 0202 14 Yes Parameter settings N 50 alpha 0 05 z 1 and M 3 Analysis performed on Mon Jun 24 16 58 39 2013 Mini Sentinel PROMPT Technical Users Guide 39 Self Controlled Design Tool Documentation egen Heure 18 SE ExampleRun Alpha Spending 7 d Target e S a 2 o Actual spent oO e O S 2 2 D o 3 ig E E E 2 23 Es C O o S e 1 2 3 1 2 3 Test Test Observed Relative Risk Log likelihood ratio
9. code and R functions that we have developed The SAS code is written exclusively for the sequential self control design while the R functions are written both for the self control design as well as for the propensity score matched concurrent controls Mini Sentinel PROMPT Technical Users Guide 1 Self Controlled Design Tool Documentation Deg This is not a guide or tutorial on the use of sequential analysis nor does it provide details about specific sequential statistical analysis or self control designs for which we refer the reader to the papers listed in the reference section After this introduction the manual is divided into four parts i study design parameter definitions ii SAS code for data extraction iii R functions for sequential analysis and iv the template for reporting the sequential analysis results Detailed step by step guidance and multiple sample screen shots are provided Il DESIGN PARAMETER SETTINGS For both the extraction of data in SAS and the analysis of data in R the user must make some decisions in advance For data extraction these include inclusion exclusion criteria age groups exposure definition health outcome definition risk and control intervals etc The data extraction SAS program is parameterized such that the user specifies these criteria in look up and mapping tables rather than in the program itself For convenience the investigators may create populate and or revise these tables
10. of two main functions AnalyzeSetUp Binomial and Analyze Binomial these functions run in tandem The first function AnalyzeSetUp Binomial establishes the pathway folder and name for the file in which the values and analyses from the extractions will be kept Each file name should be unique to both analysis parameters and a Drug and Health Outcome of Interest HOI pair Mini Sentinel PROMPT Technical Users Guide 29 Self Controlled Design Tool Documentation E iud The second function Analyze Binomial is where the user will call a filename and values from the SAS extraction table from Data Partners will be used as cases and controls for the sequential analysis 1 Function AnalyzeSetUp Binomial Prior to any analysis the user will specify a location for and name for each cumulative data file and unique parameters for all Drug HOI pairs The cumulative files will be created with AnalyzeSetUp Binomial function in the R Console window with a case sensitive template as follows AnalyzeSetUp Binomial namez N alpha z M title where name name of the file that will store cumulative analyses of one Drug HOI pair The other items to be entered into the template include the N alpha and M discussed in Section B 2 namely N number of cases in risk and control windows combined at which to stop surveillance without rejecting the null hypothesis alpha alpha level for the surveillance applied to all t
11. spent Test Cases Controls Cases Controls E Cases RR LLR target actual CV HO rejected 3 5 1 456 0 0156 0 0156 6 Parameter settings N 50 alpha 0 05 z 1 and M 3 Analysis performed on Mon Jun 24 15 30 27 2013 Target alpha Actual alpha spent spent Graphic output from R showing the numbers of cases observed expected and needed to reject the null hypothesis alpha spending observed relative risk and log likelihood ratio for the last test run will appear in a separate window within the R program Mini Sentinel PROMPT Technical Users Guide 33 Self Controlled Design Tool Documentation Figure 12 R Graphics Device 2 ACTIVE ExampleRun Alpha Spending Needed to reject HO Cv Target Observed Actual spent 4 Expected E Cases j A Cummulative Cases Alpha spending 1 1 Test Test Observed Relative Risk Log likelihood ratio Observed relative risk Observed LLR Mini Sentinel PROMPT Technical Users Guide 34 Self Controlled Design Tool Documentation El Wee D EXAMPLE TESTS 1 Analyze Binomial namez TestFile test 1 cases 5 controls lt 1 Test 1 is explained in figures 13 and 14 Figure 13 4 RGui 32 bit File Edit View Misc Packages Windows Help R Console gt Analyze Binomial name TestFile test 1 cases lt 5 controls 1 Cumulative alpha spent Test Cases Controls Cases Controls E Cases RR LLR target actual CV HO rejected 3 5 1 456
12. the drug vaccine of interest if that latter exposure is to be included A limitation of the current data extraction program is that the look back period to determine whether an exposure of interest is incident cannot differ from the look back period for the other exposure exclusions Table 9 QUERY 1 fa MULT 183 fa 2 p MULT 183 H 3 MULT 183 The health outcomes of interest are defined by diagnosis or procedure codes medical settings and whether a code or codes were the principal diagnosis In our examples all the outcomes of interest are defined by ICD9 diagnosis codes rather than CPT4 procedure codes The EVENT CODES table is populated with those codes for each of the outcomes namely angioedema SUBGROUP 101 AMI SUBGROUP 102 and seizures SUBGROUP 103 The EVENT TDL FLAG is set to 1 for codes corresponding to the health outcome of interest Settings AV ED IP can be used with diagnosis or procedure codes to define health outcomes of interest as explained below the EVENT CODES table However algorithms of greater complexity e g Dx Code A OR B ORC AND fever of X degrees OR Dx Code D AND E cannot be accommodated by the current program Table 10 EVENT CODES 101 DX09 951 1 1 102 Dxo9 410 00 102 Dxo9 410 01 102 pang 410 10 102 Dxo9 410 11 102 DX09 410 20 102 os 41021 102 Dxo9 410 30 102 DX09 410 31 PI P PI P RPITPI PITP Mini Sentinel PROMPT Techni
13. unless explicitly modified to do so Hence it should be used with caution for very young children whose baseline risk of certain health outcomes can change quickly within small increments of age such as at 3 versus 5 weeks of age It is then important to adjust for age using an offset term The self control design should also be used with caution if there is seasonality in both drug vaccine initiation and the outcome under study in which case seasonality must be adjusted for If season influences only one of these two i e either the drug vaccine exposure or the outcome the method will still be unbiased and no adjustment is needed Another potential source of bias is if there is something that triggers both the exposure and the outcome For example patients are often given the Pneumoccocal Polysaccharide Vaccine PPSV vaccine just before an organ transplant and organ transplants may cause adverse events Hence in a self control analysis there may be more adverse events just after PPSV vaccination even though the vaccine is just an innocent by stander If occurrence of the outcome influences or could influence whether the drug vaccine is given to the patient it is important to use a control window after the risk window rather than a pre exposure control window in order to avoid bias by indication or contra indication This manual describes how self control sequential analysis can be conducted within the Mini Sentinel PROMPT project utilizing SAS
14. 0 0156 0 0156 6 Parameter settings N 50 alpha 0 05 z 1 and M Analysis performed on Mon Jun 24 15 30 27 2013 Mini Sentinel PROMPT Technical Users Guide 35 Self Controlled Design Tool Documentation egen Figure 14 R Graphics Device 2 ACTIVE ExampleRun MES e Needed fo reject HO C Si o w LO e O Observed d o Expected E Cases o Eom e Go S N o c E 5 d o 4 g 1 1 Test Test Observed Relative Risk Log likelihood ratio n ae LO e D nd ZS 7 do Di S 9 CH CH c LO wn D 8 o z o 1 1 Test Test Mini Sentinel PROMPT Technical Users Guide 36 Self Controlled Design Tool Documentation 2 Analyze Binomial namez TestFile test 2 cases 3 controls 0 Test 2 is explained in figures 15 and 16 Figure 15 RGui 32 bit File Edit Yiew Misc Packages Windows Help elek elel as R Console gt Analyze Binomial name TestFile test 2 cases lt 3 controls 0 Cumulative Controls E Cases RR LLR target 456 0 0156 099 0 0156 Parameter settings N 50 alpha 0 05 z 1 and M 3 Analysis performed on Mon Jun 24 16 50 21 2013 T CH Mini Sentinel PROMPT Technical Users Guide 37 Self Controlled Design Tool Documentation egen Figure 16 R Graphics Device 2 ACTIVE ExampleRun Alpha Spending o o c Needed to reject HO o Target e _ Obstive E G Actual spent o D 5
15. 6 days 15 to 28 This is the window period either pre or post exposure Day 0 during which the risk is considered to be at baseline Maximum 35 to 28 days 1 to 28 days 1 to 28 days 1 to 28 days period to collect to include both windows Include only those with no dispensing of any drug in the drug look up table for the QUERY_GRP in question during the 0 183 days prior to the dispensing date Individuals who initiated both the drug of interest and an excluded drug on the dispensing date Day O are to be excluded Two considerations go into this look back period One is clinical for example a single episode of illness might lead to multiple visits so we specify that we want to count cases only if they represent the first occurrence of the condition in a certain number of days The lisinopril angioedema and MMRV seizures studies are examples of this where we set the look back period to 6 months 183 days because we didn t want to include repeat visits for the same episode of illness The other consideration is method related with the self control design the only informative patients are ones that have the outcome of interest in either the risk or control interval patients who have a case of the outcome in the risk interval and another case in the control interval would not be used Therefore we specify that the case must be the first in the maximum span between risk and control intervals as in
16. M iud MINI SENTINEL PROSPECTIVE ROUTINE OBSERVATIONAL MONITORING PROGRAM TOOL CONTINUOUS AND GROUP SEQUENTIAL ANALYSIS WITH SELF CONTROL DESIGN Technical Users Guide version 1 0 Prepared by Claudia Coronel Moreno MPH Yury Vilk PhD Ivair Silva PhD Katherine Yih PhD MPH Martin Kulldorff PhD Author Affiliations 1 Department of Population Medicine Harvard Medical School and Harvard Pilgrim Health Care Institute Boston MA May 19 2014 Mini Sentinel is a pilot project sponsored by the U S Food and Drug Administration FDA to inform and facilitate development of a fully operational active surveillance system the Sentinel System for monitoring the safety of FDA regulated medical products Mini Sentinel is one piece of the Sentinel Initiative a multi faceted effort by the FDA to develop a national electronic system that will complement existing methods of safety surveillance Mini Sentinel Collaborators include Data and Academic Partners that provide access to health care data and ongoing scientific technical methodological and organizational expertise The Mini Sentinel Coordinating Center is funded by the FDA through the Department of Health and Human Services HHS Contract number HHSF223200910006I E iu We Mini Sentinel Prospective Routine Observational Monitoring Program Tool Continuous and Group Sequential Analysis with Self Control Design Technical Users Guide version 1 0 Table of Contents
17. ROUP u ACEI S ARBS DIRECT RENIN INHIBITORS SUBGROUP 12 ANTIBIOTICS SUBGROUP 01 ANGIODEMA SUBGROUP 102 AMI SUBGROUP 103 ISEIZURES ENCTYPE IP IINPATIENT HOSPITAL STAYS ENCTYPE IS INON ACUTE INSTITUTIONAL STAYS ENCTYPE ED EMERGENCY DEPARTMENT ENCTYPE AV AMBULATORY VISIT ENCTYPE QA OTHER AMBULATORY VISIT PDX P PRINCIPAL DIAGNOSIS PDX NON PRINCIPAL DIAGNOSIS In the LKP_FOR_ALL_CODES table above there are three QUERY_GRPSs corresponding to lisinopril clindamycin and MMRV exposures and the conditions that define those and four EVENT_GRPs representing the health outcomes and the conditions that define those there are four instead of three EVENT_GRPs because the MMRV exposure is being studied using two different control intervals These groups are linked to each other in the GROUPS MAPPING table Mini Sentinel PROMPT Technical Users Guide 6 Self Controlled Design Tool Documentation T CH Table 4 GROUPS MAPPING GROUP QUERY GRP EVENT GRP 1 1 1 2 2 2 3 3 3 4 3 4 The minimum enrollment criteria are specified with other criteria to be discussed later in the QUERY table In these studies 183 days of continuous enrollment prior to the exposure WASHPER are required for inclusion If apparent gaps in enrollment are to be bridged ignored the macro parameter ENROLGAP should be filled out with the maximum gap to be bridged in terms of days for example 14 The default value of ENROLGAP i
18. ave only one case or control while others have more than one the functions perform a hybrid continuous group sequential analysis The R code script named Combined Analyze Binomial v1 0 was created for self controlled risk interval continuous and group sequential analysis No more than the pre defined parcel of Type error alpha is spent at each sequential test see the descriptive readme txt file included with the Combined Analyze Binomial v1 0 script The data set to be used will be from the data extraction described previously in this manual and will consist simply of counts of the health outcome of interest in a risk interval after the drug or vaccine exposure and in a control interval referred to as cases and controls respectively in the R code output To utilize the R program for sequential analysis you will open the R program open the R script Combined Analyze Binomial v1 0 run the R script Combined Analyze Binomial v1 0 enter the parameters of up to two separate functions and subsequently run the function to analyze your dataset What follows are step by step instructions to load and run the R code functions for a dataset analysis as well as explanations of the R code output A DOWNLOADING R AND LOADING FUNCTIONS The R program is available for download from the Comprehensive R Archive Network CRAN http cran r project org Detailed instructions for various operating systems are listed on the CRAN website Once you
19. cal Users Guide 9 Self Controlled Design Tool Documentation E Cm 102 xo 10401 n 102 Dm na 1 102 DX09 410 50 1 1 102 DX09 4410511 1 102 DX09 410 60 1 1 102 xo ja0611 bh 102 xo 40 70 1 H 102 Dm 410711 1 102 xo fa10 80 1 H 102 DX09 410811 1 102 DX09 410 90 1 1 102 DX09 410911 1 103 DX09 45 051 1 103 xo Das ici 103 xo X B42 H 103 Dm 345 3 1 1 103 xos 345 4 1 103 xo aussi H 103 DX09 345 6 1 1 103 DX09 345 7 1 1 103 DX09 Das gesi 1 103 DX09 Das gesi 1 103 xo 780 31 1 1 103 Dxo9 780 32 1 103 Dxo9 78039 1 1 The EVENT table specifies the medical setting s in which to identify the health outcomes of interest i e the setting s included in the algorithm whether there is a requirement that a DX or PX code be a principal diagnosis and the risk and control intervals relative to the exposure Day 0 Mini Sentinel PROMPT Technical Users Guide 10 Self Controlled Design Tool Documentation E iud Table 11 EVENT EVEN SUB ENC P WAS Es GRO pe H UP S i 35 8 1 MIN DAYS FR IND MAX DAYS FR IND MIN DAYS FR IN MAX DAYS FR IN EX FOR COMP WIEN FOR COMP WNIDEN FOR RISK W DEX FOR RISK W 102 je E 15 28 1 T 3 hos lP MUL 183 1 i 7 10 103 ED MuL 183 103 MUL 183 15 103 ED wua 15 The health outcome exclusion cri
20. chnical Users Guide 27 Self Controlled Design Tool Documentation ge Click on the newly opened window of the Combined_Analyze Binomial_v1 0 script highlight all of the script Figure 6 RGui 32 bit File Edit Packages Windows Help R version 2 15 2 2012 10 26 Trick or Treat Copyright C 20172m geg o er gt e PEL z ISBN 3 900051 07 7 4 10 Module 1 3R code Combined_Analyze Binomial R Editor Platform i386 w6 pc u X R is free softuartl Function to perform the unpredic You are welcome tc Type license Natural language AnalyzeSetUp Binomial lt function name N alpha 0 05 z 1 N 1 title R is a collaborat Type contributor citation on he ee Type demo for Bee help start fo W iee Type q to quit lt choose dir default caption past lect the folder where th m s m titlecheck lt paste name title txt sep Previously saved a Ie sr ete j amp y Temp amp y TEMP y substr address1 i i 3 1i 3 i 3 i 3 gt Ir at NA address lt addressi Once the script has been highlighted click the run ed button that appears in the Menu Bar Figure 7 R Console R version 2 15 2 2012 10 26 Trick oy Copyright C 2012 The R Foundation for St ISBN 3 900051 07 0 Mini Sentinel PROMPT Technical Users Guide 28 Self Controlled Desi
21. co x D 2 x o wo 2 j aS D o o e 7 z D iD e E a un in O e e 1 2 3 1 2 3 Test Test E EXTENDED SURVEILLANCE It is possible to continue analyzing the data even after the null hypothesis has been rejected or after the upper limit for surveillance N has been reached even though the sequential analysis has formally ended One might want to continue after rejection of the null in order to get a more stable estimate of the relative risk for example And if the null hypothesis is not rejected by the time the upper limit N is reached one might want to continue beyond the upper limit in order to gain additional assurance that the product is safe However it must be noted that no more alpha is spent and the decision of whether to reject the null hypothesis or not will not be changed Mini Sentinel PROMPT Technical Users Guide 40 Self Controlled Design Tool Documentation TT CH F REPORT FORMAT A template for reporting the results of the sequential analyses starts on the next page Currently its production is manual drawing from the outputs of the SAS data extraction programs and the R programs Mini Sentinel PROMPT Technical Users Guide 41 Self Controlled Design Tool Documentation Lisinopril Angioedema Report Format Template Lisinopril angioedema Self controlled group sequential analysis Risk interval Days 1 28 control interval Days 35 through 8 where Day 0 is the first day of exposure Sequen
22. comparison and risk windows e Create variable for number days enrolled after exposure e Set CENSORED FLAG 1 when enrollment ends within a comparison or risk window e The variable can be used to censor events e Create Final datasets e Create patient exposure episode level dataset in DPLocal library e Keep at a datapartner site amp out tbl prefix DAT EVENT e Create summary exposure event level dataset in DPLocal library Mini Sentinel PROMPT Technical Users Guide 21 Self Controlled Design Tool Documentation M iud e Keep at datapartner site amp out tbl prefix SUM EXPS EVNT LEV e Create summary dataset on event level in MSOC library e Tosend to MSOC MSOC amp out tbl prefix SUM EVNT LEV e Create summary dataset on exposure period level stratified by gender and age group in MSOC library e Tosend to MSOC MSOC amp out tbl prefix SUM EXPS PERIOD LEV e Create final summary dataset on exposure period level in MSOC library no gender and age stratification e Tosend to MSOC MSOC amp out tbl prefix SUM EXPS PERIOD LEV FIN C FOR DATA PARTNERS For the extraction of data the Data Partners will each receive a zip file which contains the SAS distributed programs and instructions for using them There are two programs for the extraction of data and the creation of output tables and a log file The values from the output tables will be used in the R code analysis Below are general instructions for using the program Mi
23. days from the EVENTFILE mapping table will specify additional conditional parameters for the identified HOI exposure pairs that fall into the risk window Utilize AGEGROUPS FILE to separate out age groups Utilize ENROLGAP macro parameter to bridge enrollment intervals separated by the grace period Use this information to further specify the cohort created in STEP 8 Create final datasets 1 local remains at Data Partner site 3 MSCDM Local Patient exposure level of every HOI in the surveillance period MSCDM Summary of exposure period level MSCDM Summary of event level MSCDM Summary of combined exposure event level Return Log to MSCDM See Appendix 1 for example of instructions sent to Data Partners IV SEQUENTIAL ANALYSIS R FUNCTIONS Data analysis will use the free and open source R statistical program The R code is written to be used both for PROMT module 1 using a self control sequential design and for PROMPT module 2 using a concurrent propensity score matched design Based on the binomial maxSPRT both use exact sequential analysis with a binomial probability model If each data feed include at most one case or control the Mini Sentinel PROMPT Technical Users Guide 23 Self Controlled Design Tool Documentation egen functions perform continuous sequential analysis If data feeds include more than one case or control the function performs a group sequential analysis with variable group sizes If soma data feeds h
24. er tete eee 40 Fe REPORT FORMA ucc RUN 41 Vi APPENDIX Do eege eege echoed gees dee ee eege eg 45 A MINI SENTINEL WORKPLAN enne nnne nnn nnnnnnnn nnn inan inin inan inii assis isis asia asia asas asia assa aas 45 EE ize eg 45 2 Overall Project Obiechive acies sseeseseeee eene tnnnnn nnns enean tn anas asses sata rasa assess essai anna nnn 45 3 MSOC Contact Information eese sees esee senses een nnne than einn as ss en tasa AE ERER 45 NEED E e EE 45 LEE reds Voie 45 6 For KP Mini Sentinel Programmers Only 46 A Workplan Timeline rr eter etr et are eter eben eee pede Ee Cove aE deve eee ed resa DEES 46 8 Package Documents Ret ere vi tees Ure ct etey set svespelie EE 46 Mini Sentinel PROMPT Technical Users Guide i Self Controlled Design Tool Documentation M iud Wee l OVERVIEW When a new vaccine or pharmaceutical drug enters the market there are always questions about its safety profile Post marketing drug and vaccine safety surveillance is important in order to detect serious adverse events that are too rare to be detected during Phase III clinical trials Such surveillance has traditionally been based on spontaneous adverse event reporting systems but electronic health records from health insurance plans are increasingly being used instead If there is a major safety problem prompt detection i
25. ests optional default 0 05 M minimum number of cases to signal optional z ratio of length of the control window to length of the risk window Lastly a title may be chosen title title for display across the tables and graphical output produced optional Once the parameters of AnalyzeSetUp Binomial function have been filled and a filename has been chosen for example name TestFile N 50 alpha 0 05 z 1 M 3 title ExampleRun gt AnalyzeSetUp Binomial name TestFile N 50 alpha 0 05 z 1 M 3 title ExampleRun the user will run the function by pressing Enter Immediately following this the user will be asked to specify the location of the file in a pop up Mini Sentinel PROMPT Technical Users Guide 30 Self Controlled Design Tool Documentation T Cm Figure 9 File Edit View Misc Packages Windows Help Browse For Folder 21 xl Select the Folder where the file TestFce C Documents and SettingslccoronellMy Documents 8H E iw inputSetU sch L My Documents pu D vri d My Computer J My Network Places set Adobe Acrobat X 1 Folder My Computer Cancel Z AnalyzeSetUp Binomial name TestFile N 50 alpha 0 05 z3 Make New Folder VVVVVV VV VV EEE EEE EE E E 2 Function Analyze Binomial The user will next be executing the function Analyze Binomial to analyze the counts of HOI events
26. gn Tool Documentation egen The Combined_Analyze Binomial_v1 0 script will begin to run in R Console window Figure 8 EI RGui 32 bit File Edit View Misc Packages Windows Help REE if sa 2 zl p lt matrix O N N 2 p i j is the probability of having j 1 cases at time mu starting probabilities are all set to zero s Calculating the first row in the p matrix for which there is a chance t foris in 1 absorb 1 p 1 s dbinom s 1 1 ps Probability of having s 1 ca p i absorb i1 i 1 pbinom absorb i1 1 1 ps f probability of rejecting HO at t for i in 2 N for j in 1 fabsorb i This loop calculates the p matix one column a for k in 1 uc i 1 1 pli 3 pli 3 p i 1 k dbinom j k 1 ps Calculates the standard p cell v for k in 1 uc i 1 1 pli absorb i 1 p i absorb i 1 p i 1 k 1 pbinom absorb i k 1 ps Calc if N gt 1 i Once the script has terminated its run you will be able to utilize the functions created for surveillance All of the functions within the R script Combined_Analyze Binomial_v1 0 are now available to load into the R Console workspace The user must run the Combined_Analyze Binomial_v1 0 script in the workspace detailed in Section D 1 Downloading R and Loading Functions at the beginning of every R session B RUNNING FUNCTIONS The R script consists
27. have downloaded the R program open it and click on the R Console window Figure 1 R Console Window 4 RGui 32 bit File Edit View Misc Packages Windows Help R version 2 15 2 2012 10 26 Trick or Treat Copyright C 2012 The R Foundation for Statistical Computing ISBN 3 900051 07 0 Platform i386 w64 mingw32 i386 32 bit R is free software and comes with ABSOLUTELY NO WARRANTY You are welcome to redistribute it under certain conditions Type license or licence for distribution details Natural language support but running in an English locale R is a collaborative project with many contributors Type contributors for more information and citation on how to cite R or R packages in publications Type demo for some demos help for on line help or help start for an HTML browser interface to help Type q to quit R Mini Sentinel PROMPT Technical Users Guide 24 Self Controlled Design Tool Documentation egen From the Menu Bar select File Figure 2 gt RGui 32 bit File Bdit View Misc Packages Windows Hel ource R code New script Open script Display filefs Load Workspace Save Workspace From the drop down of the File Menu select Open Script Figure 3 RGui 32 bit File Edit Yiew Misc Packages Windows Help Source R code Load Workspace 2012 10 26 Trick or Treat Save Workspace 12 The R F
28. ics RIT taen ii Olt Lax Lisinopril and Angioedema May 21 2013 CCM Alpha Spending Needed to reject HO CV Target S Observed Actual spent r 8 4 Expected E Cases e H D CH o c 29 O moe 2 A Ea 7 3 o 2 E a 8 E E O ce S ce Test Test Observed Relative Risk Log likelihood ratio mess co o g x 2 o Es ke R ki o o n 5 o Bow 5 e Ke T T T T T T T T T T 1 2 3 4 5 1 2 3 4 5 Test Test Mini Sentinel PROMPT Technical Users Guide 43 Self Controlled Design Tool Documentation Lisinopril Angioedema Report Format Template Data extraction history Misi Sentinel Batch Data No of No of No of Earliest Latest Date partner exposed events in events in possible possible dataset patients risk interval control dispensing dispensing created cases interval date in date in con dataset dataset trols 1 2 3 4 5 Mini Sentinel PROMPT Technical Users Guide 44 Self Controlled Design Tool Documentation ud e V APPENDIX 1 A MINI SENTINEL WORKPLAN Date submitted June 18 2013 1 Project Name Task Order 8 Foundational Elements Statistical Methods Development Mini Sentinel Request ID mini sentinel to08 scd wp1 b2 Task Order 8 Self Control Design Sequential Methods Workplan 1 Beta 2 2 Overall Project Objective The purpose of Mini Se
29. id amp out tbl prefix sum evnt sas7bdat amp dpid amp siteid amp out tbl prefix sum exps period sas7bdat amp dpid amp siteid amp out tbl prefix signature sas7bdat amp dpid amp siteid amp out tbl prefix sum exps period fin 1 LOG file returned to MSOC amp dpid amp siteid amp out rpt prefix log 1 LST file returned to MSOC amp dpid amp siteid amp out rpt prefix valid rpt Ist This program also creates 4 SAS datasets that are saved into the dplocal subdirectory These files are to remain with the Data Partner and retained for troubleshooting 4 SAS output dataset to remain at the Data Partner in the dplocal subdirectory amp dpid amp siteid amp out tbl prefix pat evnt amp dpid amp siteid amp out tbl prefix sum exps evnt amp dpid amp siteid amp out tbl prefix pat exps excl amp dpid amp siteid amp out tbl prefix pat evnt excl Mini Sentinel PROMPT Technical Users Guide 47 Self Controlled Design Tool Documentation
30. in Access but they must be imported into SAS in order for the extraction program to call them For data analysis parameters such as alpha level and maximum length of surveillance must be specified These are entered directly into the R function A DATA EXTRACTION We present the items to be specified and look up tables by means of four example studies These are 1 lisinopril and angioedema 2 clindamycin and acute myocardial infarction AMI 3 MMRV vaccine and seizures with a 1 6 day post vaccination control interval or window and 4 MMRV vaccine and seizures with a 15 28 day post vaccination control interval To extract appropriate data for analysis using this self controlled design a number of features must be specified based on clinical and or epidemiologic judgment These are listed and illustrated via the four example studies in the table below Table 1 Specified Features for Self Controlled Design LISINOPRIL CLINDAMYCIN MMRV 1 MMRV 2 NOTES Minimum period 183 days 183 days 183 days All fields same An apparent gap of continuous as for MMRV 1 in enrollment enrollment with except for can be specified pharmacy and control window such that the medical benefits gap is bridged required prior to ignored dispensing date Age range at 218 yrs 245 yrs 12 23 mo exposure Mini Sentinel PROMPT Technical Users Guide 2 Self Controlled Design Tool Documentation M iud
31. l o VOA OT 1 IL DESIGN PARAMETER SETTINGS 0 5 0 sicssscssccvscosccossvsnccovesesccsvsseccavessscorvdcescesvensscsvedsocssvesecssvenstssesesse 2 A DATA ETRINGER 2 B DATA ANALYSIS EET 5 C STATISTICAL POWER AND EXPECTED TIME TO SIGNAL sies eene enne 5 I DATA EXTRACTION SAS CODE c ccccssccccoosscscccosseseccossescoccsescecossesecoossescecesssseccosssscevossscesessescceess 5 As Tender nc 5 B FOR MINI SENTINEL OPERATIONS CENTER 12 T scole 12 2 Program E 21 C FORDATAPARINERS 3 2 2c hifi niece tecto eege 22 IV SEQUENTIAL ANALYSIS R FUNCTIONS eese nne eee eene enne nenne EE esses n hse se ssh sese ene 23 A DOWNLOADING R AND LOADING EUNCTIONS eene n nen nns sensisse nennen 24 Bz RUNNING FUNCTIONS cnini EES 29 1 Function AnalyzeSetUp Bioomiol esse eeeeeeeee eene enhn nnne nth ananas sse n eaa ra ananas sse naa 30 2 Function Z nolhvze Binomiolf essei eene eene tnn nnne nean tha nas senses satanas asas sesso sa rada n nnns 31 C ROUTPUT Soest cen AE uu LM IM E 33 PESO Dice 35 1 Analyze Binomial name TestFile test 1 CASCS lt 5 controls Jl 35 2 Analyze Binomial name TestFile test 2 CASCS lt 3 controls OI 37 3 Analyze Binomial name TestFile test 3 CASCS lt 7 controls JI 39 E EXTENDED SURVEILLANCE ooi itte etie easet ette tare eese te aci
32. n the number of cases and controls In order to run additional sequential tests using the same analysis parameters for an exposure HOI pair the user will repeat the steps in Section D 1 Downloading R and Loading Functions to define the function Analyze Binomial and then follow the steps in D 2 b Analyze Binomial using the Mini Sentinel PROMPT Technical Users Guide 32 Self Controlled Design Tool Documentation egen Analyze Binomial function and referencing the already existing file containing the stored values from previous analyses C ROUTPUT The output table in the R Console will include for each test the counts of cases and controls per test and cumulatively the expected number of cases under the null hypothesis E Cases the risk ratio RR log likelihood ratio test statistic LLR target alpha spent target actual alpha spent actual cumulative number of cases needed in the risk window in order to reject the null hypothesis critical value CV and whether the null hypothesis was rejected HO rejected Figure 11 RGui 32 bit File Edit View Misc Packages Windows Help R Console gt Analyze Binomial name TestFile test 1 cases lt 5 controls c 1 Log Likelihood Events in Events in Ratio I Reject Hull Risk Window Control Window Risk Ratio Critical Value Hypothesis Do not rejedt HO Proceed to a new test as goon aq you have more data Cumulative alpha
33. n the sequence of exposure periods specified for each data partner because the analysis method does not allow for refreshment correction and re use of previously analyzed data In the examples below HM HPHC data were obtained monthly with one additional consecutive month s worth of dispensings collected each time while chunks of dispensings for AE OS and KP NC were 3 months long Table 13 EXPOSURE PERIODS HM HPHC fa 7 1 2012 7 31 2012 mamen Ann pw euch Sum men 29900 pw eucb numm o HM HPHC je 12 1 2012 12 31 2012 AE ls Rh 1 1 2012 E ls p 4 1 2012 KP Nc Rh 1 1 2012 3 31 2012 KP NC 2 4 1 2012 If these tables are filled out in Access they must be imported into SAS B FOR MINI SENTINEL OPERATIONS CENTER 1 Documentation This documentation exists as a stand alone document called Mod prg 410 Activity Documentation docx This document has documentation for two programs in the package for Mini Sentinel activity 4 10 Mini Sentinel PROMPT Technical Users Guide 12 Self Controlled Design Tool Documentation Deen 1 Program run_mp410_distrib_prg sas This is a wrapper for running modular program mp410 In the first section of the program one needs to update parameters The rest of the program has include statements for standard MSCDM macros the main macro mp410 sas and finally the macro call mp410 2 Macro mp410 sas This is the main macro which performs the actual data manipulati
34. ni Sentinel PROMPT Technical Users Guide 22 Self Controlled Design Tool Documentation E iud Download zip file containing SAS program and mapping and lookup tables The first program is called run mp410 distrib prg sas and will extract data based on the parameters chosen It contains a macro mp4 10 sas that will perform calculations from extracted data to create variables for SAS output tables The file QUERY CODES FILE is a macro parameter that will organize the codes from Ikp for all codes sas7bdat into groups for selection of the cohort with the HOI and the exposure Join the mapping tables EVENTFILE HOI defined and QUERYFILE exposure group defined as directed by instructions in zip file A link will be established between HOI and exposure criteria Extract medical claims from MSCDM diagnosis and procedure files Extract drug claims from MSCDM outpatient and pharmacy files Combine NDC DX and PX data extracts Check the data for missing fields and set PDX as missing when it appears as a secondary or any non principal diagnosis Using the QUERYFILE and EXPOSURE PERIODS FILE mapping tables apply population selection criteria to locate all patients with exposure Utilize the EVENTFILE and INCEVENTFILE mapping tables to define the population of patients exposed those exposed patients who also were diagnosed with the HOI within the risk or control windows of exposure The WASHPER condition length of event washout period in
35. ntinel Activity 4 10 is to build the capability for active surveillance for medical product safety for FDA through approaches that are flexible enough to be easily adapted for the study of a variety of products and outcomes This data request creates analytic datasets that will be used to conduct sequential analysis using the R statistical package 3 MSOC Contact Information Name Jen Popovic Institution Mini Sentinel Operations Center Harvard Pilgrim Health Care Institute Phone 617 509 9811 Email jennifer_popovic harvardpilgrim org 4 Budget Item Task Order 8 Year 3 Foundational Elements 5 Instructions Please follow the instructions in the header section of the main SAS program mini_sentinel_to08_scd_wp1_b2 sas The sections in the header that will need user input are Edit DPID and Site ID according to the table below Edit this section to reflect your name for each Table File or View Edit this section to reflect locations for the Libraries and Folders for Mini Sentinel For this run please do NOT alter the value of LET exposure period cd 2 statement at the very beginning of the header Mini Sentinel PROMPT Technical Users Guide 45 Self Controlled Design Tool Documentation Wegen CIE When done please 1 zip compress the MSOC output folder file name should be mini sentinel toO8 scd wp1 b2 zip 2 upload it to the Mini Sentinel secure portal https portal mini sentinel org in your site s
36. nuscripts by Kulldorff et al Ill DATA EXTRACTION SAS CODE A FORINVESTIGATORS To implement the criteria and features of data to be extracted that were presented in Section B 1 one must first assign and define some codes as in the following look up table LKP FOR ALL CODES Note about terminology In the look up and mapping tables the word query is related to the exposure of interest and the word event is related to the health outcome of interest Also please note that exposure and dispensing are used somewhat interchangeably and health outcome outcome and HOI are also used somewhat interchangeably in this Section C 1 Kulldorff M Silva I R 2012 Continuous Sequential Analysis with a Delayed Start Manuscript submitted for publication Silva I R Kulldorff M 2012 Continuous versus Group Sequential Analysis for Vaccine and Drug Safety Surveillance Unpublished manuscript Mini Sentinel PROMPT Technical Users Guide 5 Self Controlled Design Tool Documentation gegen Table 3 LKP_FOR_ALL_CODES COLUMN_NAME CODE_VALUE CODE_VALUE_C CODE_DESC EVENT GRP f LISINOPRIL EVENT_GRP D CLINDAMYCIN EVENT GRP B IMMRV 1 EVENT GRP 4 IMMRV 2 GROUP 1 LISINOPRIL GROUP p CLINDAMYCIN GROUP E MMRV 1 GROUP 4 IMMRV 2 QUERY GRP f LISINOPRIL QUERY epp h2 CLINDAMYCIN QUERY GRP B IMMRV SUBGROUP 1 LISINOPRIL SUBGROUP 2 CLINDAMYCIN SUBGROUP 3 IMMRV SUBG
37. on and calculation for Mini Sentinel activity 4 10 EEEELEEAAEEALELEL ALLEL EAL EE 2K K K KK K OK kK K K Program run_mp410_distrib_prg sas Project Mini Sentinel Activity 4 10 Purpose Wrapper to run all necessary programs and macros Input parameters Standard MSCDM input parameters including parameters for libraries directories site id etc EXPOSURE_PERIOD_CD Macro parameter Should be used in conjunction with the dataset EXPOSURE EPISODES FILE See below for details Numeric 1 2 3 Required The dataset EXPOSURE EPISODES FILE is distributed as a part of the package In addition to the column EXPOSURE PERIOD CD the dataset has two columns EXPOS PERIOD STRT DT and EXPOS PERIOD END DT These dates can be different for different data partner sites depending on sites data availability and refresh rate For example the exposure period can be a month length for one site and a quarter for another This is possible because the study compares two statistics calculated at same data partner site the number of adverse outcomes in risk time window with the number of adverse outcomes in comparison time window The time periods appropriate to the site should be supplied by MINI SENTINEL OPERATIONS CENTER or updated by local site per leading site instructions The allowed values are either a single number or a list of numbers separated by spaces The latter is mostly used on historical data to determine background rates E
38. oundation for Statistical Computing 0 Load History 64 minqw32 i386 32 bit Save History Mini Sentinel PROMPT Technical Users Guide 25 Self Controlled Design Tool Documentation TU CORN A pop up window will ask the user to locate the R script The R script is distributed separately from the User Manual and is named Combined Analyze Binomial v1 0 The user is responsible for deciding where this script is saved Figure 4 SA RGui 32 bit File Edit View Misc Packages Windows Help 5 e 3 85 o Look in Mee Add in Express Downloads My Music My Pictures My SAS Files Files of type R files Ri Mini Sentinel PROMPT Technical Users Guide 26 Self Controlled Design Tool Documentation Select Combined_Analyze Binomial_v1 0 and open this file Figure 5 RGui 32 bit File Edit View Misc Packages Windows Help elelee Open script 2 x Look in O R code DI orf EN H Code binomial surveillance Conservative V13 R KE Code binomial surveillance Conservative 13 R 2013 06 04 V 14 CCP E Code binomial surveillance Conservative 14 R d wordDoc f Code binomial surveillance Conservative V14 R 2013 6 15 id wordDoc KE Code binomial surveillance Conservative V14 R 2013 6 19 iM wordDoc Bic Code noma slane Conservative V14 R 2013 06 25 M WordDoct Speer vativeV14 R 2013 6 24 ja Files of type aes zl Cancel E A Mini Sentinel PROMPT Te
39. que per MSCDM data model e The last two steps are added in case a site does not completely comply with MSCDM data model e Separate extracted encounters into ones that will be used for population selection query conditions and the ones that will be used for adverse outcome events conditions e Reset PDX to missing when it is not principal diagnosis because we need to treat missing and secondary diagnosis in the same way e Add population selection parameters from the mapping table to the QUERY INQUERY data extract e Apply population selection criteria to find patients with exposure to the drugs e or other encounters of interest e Extract enrollment and demographics information for the population of interest e Create enrollment spans which bridge enrollment intervals separated by allowed grace period e Add enrollment and demographics data to the table with patients who had exposures e Calculate age in units years months etc defined by parameter AGE INTERVAL TYPE e Exclude patients that do not have age in the required age intervals e Find patients that can have outcomes of interest and add parameters for adverse event conditions e Find patients who satisfy WASHPER condition for adverse events e Combine exposure information and adverse outcome information into a single dataset e Match query conditions QUERY GRP with the corresponding event conditions EVEN GRP e Add study group code e Create flags for events in
40. s 45 Table 5 QUERY QUERY_GRP SUBGROUP WASHTYP WASHPER EXCL_COMBO_MEDS_ON_SAME_DAY 1 1 MULT 183 2 2 MULT 183 3 3 MULT 183 The age range is specified in the AGE GROUPS table below which also specifies and assigns codes to the age groups within the range The age groups within the ranges are not needed for the self controlled analysis but are used to characterize cases in the event that a signal appears that warrants further investigation Table 6 AGE GROUPS Y 18 44 w w NINININI N ee ee lt UI D On E N e olol Nuala N oaan e The NDC codes identifying the exposures of interest lisinopril clindamycin and MMRV and the kinds of prior exposures for which exposures of interest are to be excluded referred to as exposure exclusions in Section B 1 are in the QUERY_CODES table This table is very long and is shown only schematically below with just one row per SUBGROUP The CODEs associated with SUBGROUPSs 1 2 and 3 are the Mini Sentinel PROMPT Technical Users Guide 7 Self Controlled Design Tool Documentation Deg 11 digit NDC codes for lisinopril clindamycin and MMRV products respectively SUBGROUPs 11 and 12 are discussed below the table in the context of other exposures for which an exposure of interest should be excluded Table 7 QUERY_CODES SUBGROUP CODETYPE CODE QUERY_TBL INCQUERY_1 1 Rx11 multiple 11 digit 1 0 numeric NDC codes 2 R
41. s critical Electronic health data can now be accessed on a quarterly monthly or even weekly basis to prospectively monitor the safety of a newly approved drug or vaccine using gradually accumulating data This repeated monitoring requires use of sequential statistical analysis which adjusts for the multiple testing inherent in the many looks at the data Self control methods are a powerful and often used design for drug and vaccine safety studies when looking for acute adverse events that occur soon after initial exposure Through the self control design all non time varying confounders are automatically adjusted for including gender insurance coverage education income geography etc The method uses a pre defined risk window such as 1 21 days after initial exposure and a pre defined control window of the same or different length Depending on the drug vaccine event pair under study the control window may be either before exposure or after the risk window Technically it is also possible to have a control window immediately after the initial exposure if an increased risk is implausible during that period and a risk window after the control window but this would not be a common study design In the current version the ratio of the risk to control window must be the same for all patients Future versions of the system will allow for different ratios for different patients The self control method does not adjust for time varying confounders
42. s in code values are allowed One can use either UNIX style wild cards or SQL style wild cards _ The presence or absence of a period in PX or DX code does not affect the results CODETYPE Details type of each code value included in the CODE field above Character 4 Required Valid values are RX11 11 digit NDC e RX09 9 digit NDC PX09 ICD 9 CM procedure PX10 ICD 10 CM procedure PX11 ICD 11 CM procedure e PXC4 CPT 4 procedure i e HCPCS Level I e PXHC HCPCS procedure e HCPCS Level PXH3 HCPCS Level Ill PXC2 CPT Category Il PXC3 CPT Category III QUERY TBL FLAG INCQUERY TBL FLAG Optional fields that identify if SUBGROUP is QUERY and or INCQUERY table They have no affect on the program but can be useful for validation EVENTFILE Macro parameter mom Mapping table for adverse events table by EVENT GRP The tables EVENTFILE and INCEVENTFILE allow NULL values that must be interpreted as any allowed values For example ENCTYPEIS NULL in one of these tables means that all five possible values of ENCTYPE are allowed The JOIN to the table LKP FOR ALL CODES create a table which has explicit set of all possible codes Mini Sentinel PROMPT Technical Users Guide 17 Self Controlled Design Tool Documentation M iud COLUMNS EVENT GRP A code which identifies set of conditions for adverse events of interest Numeric 1 2 3 Required SUBGROUP Identifies group of codes typicall
43. space after the left pointing arrow controls numeric value from extraction table of the HOI events that occur in the control window again when entering the number leave one space after the left pointing arrow As an example we have chosen the following values to run our analysis name TestFile test 1 cases 5 controls lt and the Analyze Binomial input for the R Console is as follows gt Analyze Binomial name TestFile test 1 cases lt 5 controls lt 1 It is important to note that for Analyze Binomial as time goes on and new data are received for analysis all values for test must be entered in chronological order in the order the corresponding data were received If any incorrect values for a test are entered then it will be necessary to delete the corresponding text file established previously via AnalysisSetUp Binomial and run all tests again The reason for deleting the file is that the values from all previous tests are stored and used and the file and any stored analyses and descriptive statistics will no longer be correct It is strongly suggested that the user keep a separate detailed log of all data extractions from contributing Data Partners After entering the values into the R Console press the Enter key to run the function Analyze Binomial The Analyze Binomial function will run sequential analysis and produce an output of a table and four graphs This may take several minutes depending o
44. teria are specified in the EVENT_CODES EVENT and INCEVENT tables As is the case for the exposure of interest there are two kinds of HOI exclusions one is if the HOI is not incident the other is if the HOI was preceded by a certain other kind of outcome within a specified period of time For excluding non incident cases we first set the look back period In our example studies we defined incident HOI as the first of its kind to occur in a 183 day period this is specified by setting WASHPER 183 in the EVENT table which leads to exclusion of an HOI occurring lt 183 days after a previous one for the same SUBGROUP Where the control interval also called the comparison window or COMP_WND is prior to exposure the look back will go back farther than the minimum enrollment specified in the QUERY table We also need to specify the setting and principal diagnosis criteria for prior cases This is done using the ENCTYPE and PDX fields in the INCEVENT table For example in the clindamycin AMI study EVENT GRP 2 we want to look for prior AMI in the inpatient IP setting only and if any such case is found in the 28 days prior to a case being considered the more recent case is excluded ENCTYPE is left blank if all three settings are to be used For example in the MMRV seizure studies EVENT GRPs 3 and 4 the seizure outcomes to be used must be in either the ED or inpatient setting but must not follow within 183 days a seizure in any setting
45. the clindamycin AMI study in which we specified a look back period of 28 days Mini Sentinel PROMPT Technical Users Guide 4 Self Controlled Design Tool Documentation M iud B DATA ANALYSIS Before analyzing the data the user must specify several parameters Table 2 Several Parameters to Consider Before Data Analysis PARAMETER EXPLANATION CONSIDERATION 0 05 is the default if no other value alpha Desired alpha level overall is chosen 1is the default which means that the system will signal as soon as M Minimum number of cases in risk there are enough events in the risk control intervals to signal interval to reject the null hypothesis It does not mean that it will signal after one event Cumulative number of cases in risk control intervals at which surveillance is to stop even if Ho has not been rejected sometimes called the upper limit on the length of surveillance No default should be selected to ensure that there is enough statistical power to detect the lowest risk of concern for public health C STATISTICAL POWER AND EXPECTED TIME TO SIGNAL Outside of this project R code has been developed to estimate a the statistical power to detect a range of relative risks b for different maximum lengths of surveillance as well as c the expected time to signal when the null hypothesis is rejected The statistical power can also be looked up in reference tables available in the ma
46. tial analysis history and results Test no and report date Last exposure date Cumulative exposed NT CHR Tracking New events Cumulative Risk estimates Hypothesis testing statistics and results Test Most No of No of No of No of No of Expec Relative Risk Log Target Actual No of HO H recent new new exposed events events ted no risk differ likeli alpha to alpha cases rejected batch es events events patients in risk in of cases ence hood spend spent needed included in risk in interval control ratio to reject interval control cases interval test HO CV cases interval con statistic con trols trols 1 2 3 4 5 Date of last analysis 5 21 2013 Sequential analysis parameter settings Performance metrics Maximum total no of events in risk and control intervals N 30 Relative risk 1 0 1 5 20 30 5 0 10 Alpha level 0 05 Estimated statistical power Ratio of risk interval to control interval z control risk 1 1 Expected total no of events when HO rejected Minimum total no of events to signal M 3 Expected total no of events at end of surveillance Mini Sentinel PROMPT Technical Users Guide 42 Self Controlled Design Tool Documentation Lisinopril Angioedema Report Format Template N Gei me Graphical representation of subset of results and statist
47. uct within a certain period of time is to lead to exclusion one first creates a separate SUBGROUP of the product codes in the QUERY CODES table For example SUBGROUP 11 consists of ACEi s ARBS and direct renin inhibitors all of which are to lead to exclusion from Study GROUP 1 if a dispensing of any of them appeared in the 183 days prior to the lisinopril exposure being considered and SUBGROUP 12 consists of antibiotics which are to lead to exclusion from Study GROUP 2 if a dispensing of any of them appeared in the 183 days prior to the clindamycin exposure being considered The INCQUERY table links the exposure exclusions SUBGROUPSs to their respective exposures of interest QUERY GRPs Mini Sentinel PROMPT Technical Users Guide 8 Self Controlled Design Tool Documentation egen Table 8 INC_QUERY QUERY_GRP SUBGROUP 1 Finally in the QUERY table which we present again below the field EXCL_ MEDS ON SAME DAY is for the purpose of indicating whether Day O the day of the exposure of interest is to be included in the look back period for exclusion The value should be set to 1 if receipt of any product in the exclusion list on the same day as the exposure of interest Day 0 within the respective QUERY GRP disqualifies a dispensing from inclusion Setting the value to 1 means that no product in the exposure exclusion list and no other formulation of the drug vaccine of interest can be given on the same day as
48. ut period in days Possible values the same as in MP3 EXCL_MEDS_ON_SAME_DAY When this flag is set to 1 then cases with more than one NDC from the same QUERY_GRP and on the same day are excluded In other words the look back period starts from O days i e look back period 0 WASHPER When this flag is set to O then the look back period starts with previous day i e look back period is 1 WASHPER Possible values are O and 1 INCQUERYFILE Macro parameter Mapping table with additional parameters for population selection Identifies conditions for previous incidence of exposure which are used to exclude exposure encounters define by QUERYFILE Used in conjunction with WASHTYP and WASHPER from QUERYFILE Optional COLUMNS QUERY GRP A code which identifies set of conditions for population selection Usually defines population who had exposures to a set of drugs Numeric 1 2 3 Required SUBGROUP Identifies group of codes typically a set of drugs Numeric 11 12 13 Required QUERY_CODES_FILE Macro parameter Mini Sentinel PROMPT Technical Users Guide 16 Self Controlled Design Tool Documentation E iud Mapping table between group of codes SUBGROUP and actual codes used for population selection QUERY INCQUERY COLUMNS SUBGROUP Identifies group of codes typically a set of drugs Numeric 11 12 13 Required CODE NDC diagnosis or procedure code mapped to a subgroup above The wild card
49. x11 ditto 1 0 3 Rx11 ditto 1 0 11 Rx11 ditto 0 1 12 Rx11 ditto 0 1 The specifications for exposure exclusion criteria appear in multiple tables the QUERY and QUERY CODES tables discussed above and the INCQUERY table In the example studies the minimum enrollment period is the same as the look back period to check for prior exposures thus the QUERY table serves both purposes if investigators wish to specify distinct minimum enrollment and look back periods an additional parameter will be needed for minimum enrollment The QUERY CODES table contains not only the codes for the exposures QUERY GRPs being studied but also codes for types of exposures meant to lead to the exclusion of an exposure of interest These two categories of product are distinguished by the QUERY TBL and INCQUERY 1 flags the first of which is set to 1 for the exposures being studied and the second of which is set to 1 for the exposures leading to exclusion There are two kinds of exposure exclusions one is if the exposure of interest is not incident the other is if the exposure of interest was preceded by exposure to some other product within a certain period of time In our example studies we defined incident exposure as the first dispensing in a 183 day period this is specified by setting WASHPER 183 in the QUERY table which leads to exclusion of a dispensing occurring 183 days after a previous one for the same SUBGROUP If exposure to some other prod
50. xamples LET exposure period cd 1 LET exposure period cd 1 2 3 Mini Sentinel PROMPT Technical Users Guide 13 Self Controlled Design Tool Documentation M iud PROGRAM mp410 PURPOSE Create summary tables for drug of interests and adverse events INPUT Patients level tables Tables in MSCDM format Parameters Lookup and Mapping tables used by the program The Lookup and Mapping tables are distributed as a part of a package for modular program EES EXPOSURE PERIOD CD Macro parameter Should be used in conjunction with the dataset EXPOSURE EPISODES See below for details Numeric 1 2 3 Required E REESE RNC REESE PRICE OO SO RIE EE CES PRECII OCR PRE SPRUE OY EET ST ENROLGAP Macro parameter Grace period used to bridge intervals of enrollment Numeric Default 45 days The dates QUERYFROM and QUERYTO below should be set very wide The actual restrictions by dates are set by the mapping tables EXPOSURE_PERIODS_FILE QUERYFILE EVENTFILE WASHPER The only case when the above dates will be used is when WASTYP MIN which rarely necessary OU a a a a IE E AAE EAER ee a Pale ah EA a Seta cae a ets aetna att TMP Macro Parameter for temporary library Default is WORK It can be changed to the permanent library for troubleshooting EE A EEEIEE AE A EA OEA PEE A OAE EAIN EAE TEANA TEIE EEEE TMP2 Macro Parameter for temporary library Default is WORK This parameter should be changed only if your
51. y a set of diagnosis Numeric 101 102 Required WASHTYP Selects how incidence events will be defined Character 3 Possible values the same as in MP3 MIN MULT WASHPER Length of event washout period in days Possible values the same as in MP3 Numeric Required 0 must be entered if no WASHPER is required ENCTYPE Encounter type The same as in MSCDM data model PDX Principal diagnosis As in MSCDM in data model it is defined only for inpatient stays Columns for comparison and risk windows MAX_DAYS_FR_INDEX_FOR_COMP_WND MAX_DAYS_FR_INDEX_FOR_RISK_WND MIN_DAYS_FR_INDEX_FOR_COMP_WND MIN_DAYS_FR_INDEX_FOR_RISK_WND INCEVENTFILE Macro parameter Mapping table with additional parameters for adverse outcomes selection Identifies conditions for previous outcomes which are used to exclude adverse outcomes defined by EVENTFILE Used in conjunction with WASHTYP and WASHPER from EVENTFILE Optional COLUMNS Same meaning and data types as in EVENTFILE EVENT_GRP SUBGROUP ENCTYPE PDX EVENT_CODES FILE Macro parameter Mapping table between group of codes SUBGROUP and actual codes used for adverse events EVENTS INCEVENTS Has similar structure as QUERY_CODES FILE The two tables are separated for convenience Mini Sentinel PROMPT Technical Users Guide 18 Self Controlled Design Tool Documentation M iud The wild cards in code values are allowed One can use either UNIX style wild cards or SQL st
52. yle wild cards _ The presence or absence of a period in PX or DX code does not affect the results AGE GROUPS FILE Macro parameter Mapping table for age groups COLUMNS QUERY GRP See QUERYFILE above AGE GROUP Code for age group Numeric 1 2 3 Required AGE UNIT Unit for measuring age Character 1 Valid values D days W weeks Q quarters e M months e Y years default value AGE_MAX Minimum age at exposure index date Numeric Default 0 AGE_MIN Maximum age at exposure index date Numeric Default 200 OUTPUT All output tables reports and logs use output prefix below out_tbl_prefix amp DPID amp REQUESTID out_rpt_prefix amp out_tbl_prefix _ amp DateStamp amp TimeStamp Summary datasets on event date relative to exposure dt level SUM EVNT Summarization level GROUP EXPOSURE PERIOD CD EXPOS PERIOD STRT DT AGE GROUP SEX DAYS FROM EXPOS TO EVENT DAYS ENR AFTER EXPOS Statistics Mini Sentinel PROMPT Technical Users Guide 19 Self Controlled Design Tool Documentation T C CNT EXP EPISODES CNT IN COMP WND CNT IN RISK WND CNT CENSORED EVENTS SUM EXPS PERIOD Summary datasets for adverse events on exposure period summarization level stratified by gender and age group Summarization level GROUP EXPOSURE PERIOD CD EXPOS PERIOD STRT DT AGE GROUP SEX Statistics CNT EXP EPISODES CNT IN COMP WND CNT IN RISK WND CNT CENSORED EVENTS MIN
Download Pdf Manuals
Related Search