PharmaSUG 2017 - Paper DS19 Standardized, Custmized r Bth? Defining and Implementing (MedDRA) Queries in ADaM Data Sets Richann Watsn, Experis, Batavia, OH ABSTRACT Karl Miller, inventiv Health, Lincln, NE Investigatin f drug safety issues fr clinical develpment will cnsistently revlve arund the experience and impact f imprtant medical ccurrences thrughut the cnduct f a clinical trial. As a first step in the data analysis prcess, Standardized MedDRA Queries (SMQs), a unique feature f MedDRA, prvide a cnsistent and efficient structure t supprt safety analysis, reprting, and als address imprtant tpics fr regulatry and industry users. A variance in wrking with SMQs is the ability t limit the scpe fr the analysis need (e.g., Brad r Narrw ) but there is als the ability utside f the specific SMQs in allwing the ability t develp Custmized Queries (CQs). With the intrductin f the ADaM Occurrence Data Structure (OCCDS) standard structure, the incrpratin f these SMQs, alng with ptential CQs, slidified the need fr cnsistent implementatin, nt nly acrss studies, but acrss drug cmpunds and even within a cmpany itself. Wrking with SMQs ne may have numerus questins: What differentiates the SMQ frm a CQ and which ne shuld be used? Are there any ther cnsideratins in implementatin f the OCCDS standards? Where des ne begin? Right here INTRODUCTION In wrking with clinical trials and the data cllected during study cnduct, ne f the mst cmmn aspects r areas that yu will encunter in handling the safety data used in analysis will cnsistently be fcused arund the Adverse Events. One aspect, and the fcus f this paper, is wrking with SMQs, which are pre-determined, validated sets f MedDRA terms, and the CQs which cnsist f specific terms f interest t the analysis. Used t grup MedDRA terms, t help define a specific medical cnditin, r define a particular area f interest SMQs and CQs are lists typically cnsisting f like Preferred Terms (PTs) develped t aid in the efficient identificatin f safety data key t analyze. Currently there exist ver 100 SMQs with mre in future develpment. Belw is a list f sme SMQs that are available: Anaphylactic reactin Cardiac failure Depressin and suicide/self-injury Hepatic disrders Hypersensitivity Ischaemic heart disease Pregnancy and nenatal tpics Rhabdmylysis/mypathy Systemic lupus erythematsus Like SMQs, CQs are used t grup like adverse events, hwever, CQs are custmized based n study need and therefre d nt have a pre-defined list that can be referenced. SPECIAL FEATURES OF SMQ Since SMQs cnsist f sets f PTs, t avid cnflict with existing MedDRA terms, each SMQ name has (SMQ) appended at the end f the query. Fr example, a specific query lking fr adverse events that 1
wuld be classified under the SMQ f Anaphylactic reactin wuld have Anaphylactic reactin (SMQ). Additinally, each SMQ is assciated with a specific SMQ cde that is 8 digits that starts with 2. Other special features, which we will discuss later, cnsist f useful ptins such as a scpe search ( Narrw vs Brad ) and algrithmic searches using class r hierarchy structure. Thrugh the use and implementatin f these special features, this will allw users t further refine the SMQ list t ptentially get the desired results withut having t create a CQ. SMQ SCOPE Within an SMQ a PT can be classified as having a Narrw r Brad scpe. The Narrw scpe identify the PTs that are mre likely t represent the cnditin r area f interest, where the Brad scpe PTs may end up having little t minimal interest fr use in the analysis upn further investigatin. Thus, if a specificity analysis is needed then the PTs within the Narrw scpe wuld be the terms f interest. Hwever, if a sensitivity analysis is needed, then all PTs assciated with the SMQ shuld be selected (i.e., PTs having either Narrw r Brad scpe). Figure 1 illustrates the cncept f specificity versus sensitivity analysis thrugh the use f Narrw and/r Brad scpe fund with the SMQs. Specificity Analysis NARROW TERMS BROAD TERMS Sensitivity Analysis Figure 1 Specificity vs. Sensitivity Analysis in Relatin t SMQ Scpe SMQ CLASS T allw fr an algrithmic search, each SMQ is further classified int classes (A, B, C r D). Fr example, there culd be sme instances where a SMQ may retrieve mre terms than what is actually needed fr a specificity analysis. In this case, the SMQ can then be further refined by selecting nly the PTs that have CLASS = B. Table 1 prvides a prtin f the SMQ table fr Acute Pancreatitis (SMQ). In this table yu can see that fr the terms with Brad scpe that they are further classified int CLASS B and C. The terms assciated with CLASS B are terms that are assciated with labratry values and the terms assciated with CLASS C are the nes assciated with signs and symptms. Thus, if an analysis fr Acute Pancreatitis is needed, dealing with labratry results nly, then the SMQ cde can be refined even further just t lk at the SCOPE = BROAD and the CLASS = B. SMQ SMQCODE SCOPE CLASS PT PTCODE Acute pancreatitis (SMQ) 20000022 Brad B Amylase abnrmal 10072327 Acute pancreatitis (SMQ) 20000022 Brad B Amylase creatinine clearance rati 10073699 abnrmal Acute pancreatitis (SMQ) 20000022 Brad B Amylase increased 10002016 Acute pancreatitis (SMQ) 20000022 Brad B Bilirubin cnjugated abnrmal 10067718 Acute pancreatitis (SMQ) 20000022 Brad B Bld bilirubin increased 10005364 Acute pancreatitis (SMQ) 20000022 Brad B Bld trypsin increased 10064751 Acute pancreatitis (SMQ) 20000022 Brad C Abdminal pain 10000081 Acute pancreatitis (SMQ) 20000022 Brad C Abdminal tenderness 10000097 Acute pancreatitis (SMQ) 20000022 Brad C Acute abdmen 10000647 Acute pancreatitis (SMQ) 20000022 Brad C Gastrintestinal pain 10017999 Acute pancreatitis (SMQ) 20000022 Brad C Nausea 10028813 Acute pancreatitis (SMQ) 20000022 Brad C Vmiting 10047700 Acute pancreatitis (SMQ) 20000022 Brad C Vmiting prjectile 10047708 2
Table 1 Brad Terms Split int Class B and Class C SMQ HIERARCHY In additin t the scpe and class feature, sme SMQs may be brken dwn further int sub-queries. These sub-smqs are related t anther SMQ in a hierarchical nature, similar t the hierarchy f MedDRA, and can help refine the search t increase specificity f the selected terms fr use in the analysis. The sub-smqs can als be used t retrieve nly thse PTs f interest but be n alert, there can be up t fur sub-queries within an SMQ. Fr example, Depressin and suicide/self-injury (SMQ) nly has ne sub-level (Figure 2) while Hepatic disrders (SMQ) has fur sub-levels (Figure 3). Figure 2 Depressin and suicide/self-injury (SMQ) Hierarchy Figure 3 Hepatic disrders (SMQ) Hierarchy 3
DIFFERENCE BETWEEN SMQ AND CQ In wrking with queries, thrugh defining a set f cnditins r simply gruping terms when applicable, there are differences in hw t wrk with SMQs and CQs. One main difference between deciding t wrk with SMQs and CQs is that the SMQs are pre-defined lists that are can be extracted using MedDRA PTs, where the CQs are lists that are specific t a study and are nt pre-defined within MedDRA. In wrking with a set, pre-defined list versus a study-by-study base, this can have an impact n the efficiency f implementatin thrugh prgramming and the ptential use f the query fr future studies. Anther difference (and yet als a benefit t bth) is that there are specific features built int the SMQ MedDRA dictinary that will allw fr sub-setting and refining the list f PTs f interest. In the instance where yu d implement the list f these pre-defined terms alng with the pre-defined features, but the resulting list f PTs is nt exactly what is needed fr analysis r reprting purpses, then this is the case where the CQs cme in handy by being able t expand the list f terms t encmpass all the PTs f interest r by further sub-setting based n additinal infrmatin. Lastly, and key t implementatin f CDISC standards, is the difference between the use f SMQ and CQ when it cmes t implementing them int an ADaM data set fr analysis and presentatin purpses. As shwn in Table 2, fr SMQs there are fur variables that can be used and therefre implemented int the specificatins and data set, while at the same time fr CQs there is nly ne variable. Variable Name Variable Label Type Cre SMQzzNAM SMQ zz Name Char Cnd SMQs SMQzzCD SMQ zz Cde Num Perm SMQzzSC SMQ zz Scpe Char Cnd SMQzzSCN SMQ zz Scpe (N) Num Perm CQs CQzzNAM Custmized Query zz Name Char Cnd Table 2 SMQ and CQ Variables Dn t let the number differences between the variables used fl yu n which will be mre efficient t implement in yur ADaM data set. In lking at the list f variables fr queries, this can be the case where less is nt always mre; just because the use f a CQ requires ne variable and SMQs can have up t at least fur, des nt always translate int the best ptin. Fr example, with SMQs the majrity f the wrk is dne up-frnt, with the pre-defined lists and implementatin methds, s there is n need fr the creatin f special lk-up tables and custm prgramming in implementatin. IMPLEMENTATION OF SMQ AND CQ VARIABLES SMQ VARIABLES When generating the specificatins fr the required analysis and implementatin f SMQs, it is always gd t keep in mind that each SMQ yu are using will have its wn set f variables. Mre than ne SMQ cannt be stred within r under the same SMQzzNAM, SMQzzCD, SMQzzSC variables in yur data set. See Table 3 fr an example f hw SMQ variables shuld be implemented int the data when there is mre than ne SMQ. One item f nte is that it is nt a requirement fr zz t be sequential when implementing multiple SMQs. Rw AEDECOD SMQ11NAM SMQ11CD SMQ11SC 1 Acute fatty liver f pregnancy Hepatic disrders (SMQ) 20000005 Narrw 2 Acute hepatitis B Hepatic disrders (SMQ) 20000005 Narrw 3 Acute n chrnic liver failure Hepatic disrders (SMQ) 20000005 Narrw 4 Agitatin nenatal 5 Vmiting 6 Ascites Hepatic disrders (SMQ) 20000005 Narrw 7 Chlestasis f pregnancy Hepatic disrders (SMQ) 20000005 Narrw 8 Nenatal chlestasis Hepatic disrders (SMQ) 20000005 Narrw 9 Jaundice Hepatic disrders (SMQ) 20000005 Narrw 4
Rw SMQ20NAM SMQ20CD SMQ20SC SMQ23NAM SMQ23CD SMQ23SC 1 Pregnancy and nenatal tpics (SMQ) 20000185 Narrw 2 3 4 Pregnancy and nenatal tpics (SMQ) 20000185 Narrw 5 Acute pancreatitis (SMQ) 20000022 Brad 6 Acute pancreatitis (SMQ) 20000022 Brad 7 Pregnancy and nenatal tpics (SMQ) 20000185 Narrw 8 Pregnancy and nenatal tpics (SMQ) 20000185 Narrw 9 Acute pancreatitis (SMQ) 20000022 Brad Table 3 Illustratin f Implementatin f SMQ Variables In the example, the zz values where purpsely set t varius numbers t illustrate that the numbering des nt have t start at 01 and des nt have t be sequential. This can be seen thrugh the use f pre-defined standard specificatins acrss an entire cmpund, r even an entire therapeutic area, where specific SMQs can be assigned t their wn variable naming cnventin acrss all clinical trials within a cmpany. CQ VARIABLE Similar t SMQ, each CQ will have its wn variable when implementing them int the ADaM data set. Just as with SMQs, mre than ne CQ cannt be stred within the same CQzzNAM variable within a data set f terms. In the case where a specific term may be included fr multiple custm lists f PTs that are f interest fr analysis, this helps keep the lists clear in their use and purpse fr the analysis utputs. HOW TO DETERMINE A CQ With queries fcused arund the PTs f clinical events, where is the line between whether a query shuld be cnsidered a SMQ r nt? SMQs grup terms and SOCs related t a defined medical cnditin r area f interest that can assist in case identificatin, thrugh many related signs, symptms, and ther findings specifically related t the medical cnditin r area f interest. Hwever, being stringent and rigid, SMQs d nt cver all medical tpics r safety issues that may be f interest, and will likely evlve and need t be refined as future prgress is made. With CQs there is mre leniency and rm t wrk utside f the defined lists f SMQs, hw can yu determine when yu re using a CQ? CQs can be determined in several different ways: Prvided list A list that is built frm ne r mre SMQs that are mdified r jined tgether t prvide a lkup table A list f PTs, typically prvided by a clinician, that are nt already defined in an SMQ Prgrammatically determined Criteria can be a SMQ with specific PTs remved that culd nt be remved using ne f the built in features Criteria prvided by statistician, clinician r in Statistical Analysis Plan (SAP) Field captured n the case reprt frm (CRF) t indicate the term/event is f special interest The fllwing examples illustrate the varius way CQs can be determined. In Table 4 we illustrate a sample lkup table that is prvided by either a clinician r statistician in the frm f a spreadsheet. This lkup table can then be cnverted t a SAS data set and then used t assign the crrect CQzzNAM value as illustrated in Table 5. 5
Rw Criteria Terms Needed 1 Anaphylaxis per Criteria 1 Anaphylactic reactin 2 Anaphylaxis per Criteria 1 Dyspnea 3 Anaphylaxis per Criteria 1 Angiedema 4 Anaphylaxis per Criteria 1 Migraine 5 Anaphylaxis per Criteria 1 Presyncpe 6 Anaphylaxis per Criteria 1 Urticaria 7 Anaphylaxis per Criteria 2 Angiedema 8 Anaphylaxis per Criteria 2 Hyperhidrsis Table 4 Sample Custmized Query Lkup Table Prvided as a List Rw AEDECOD CQ01NAM CQ02NAM 1 Anaphylactic reactin Anaphylaxis per Criteria 1 2 Angiedema Anaphylaxis per Criteria 1 Anaphylaxis per Criteria 2 3 Liver functin test abnrmal 4 Migraine Anaphylaxis per Criteria 1 5 Injectin site jint pain 6 Injectin site erythema Table 5 Illustratin f Implementatin f CQ Variables Using a CQ Lkup Table In additin t using lkup tables, CQs can be prgrammatically determined based n prvided criteria frm a clinician, the SAP, r directly in the specificatins fr the CQ. As illustrated in Table 6, assume that the specificatins stated, under the derivatin fr the CQ variable (CQ11NAM), that any AE wuld be marked as a term f interest in the analysis data if AEBODSYS = Infectins and infestatins and AESER = Y. Rw USUBJID AEBODSYS AEDECOD AESER CQ11NAM 1 ABC-DEF-1234 Infectins and infestatins Laryngitis N 2 ABC-DEF-1234 Infectins and infestatins Sinusitis Y Serius Infectins 3 ABC-DEF-1234 Nervus system disrders Headache N 4 ABC-DEF-1234 Nervus system disrders Migraine Y 5 ABC-DEF-1234 Skin and subcutaneus tissue disrders Angiedema N 6 ABC-DEF-5678 Investigatins Liver functin test abnrmal N 7 ABC-DEF-5678 Infectins and infestatins Naspharyngitis Y Serius Infectins 8 ABC-DEF-5678 Jaundice Sinusitis N Table 6 Illustratin f Implementatin f CQ Variable Determined Prgrammatically Furthermre, CQs can be determined based n field captured n the CRF. Display 1, shws an example CRF fr the Adverse Event frm. On the frm there is a field t indicate whether the AE is f special interest t the principal investigatr. This field, r ne similar t it, can be used t define the CQzzNAM variable lgic fund in the study specificatins. Display 1 Sample CRF f an Adverse Event Frm 6
GOOD IMPLEMENTATION CONSIDERATIONS With the availability fr use with ADaM OCCDS v.1.0, all implementatin methds frm this paper can be implemented fr creating the analysis data sets. With the many ptins, it s always gd t keep certain scenaris in mind when implementatin is set t start: Cannt stre multiple cdes in the same variable fr either SMQs r CQs At the cmpund, r even therapeutic area level, yu will want t keep the naming f the SMQ and CQ variables cnsistent t assist in ptential integratin at a later time CQs will require the client t be respnsible fr any MedDRA versin updates/changes Be aware f the ptential change in SMQs thrugh different versins f MedDRA All terms included in a SMQ will have a status s different versins may have a term inactivated ver time, but nce a term is added t a SMQ it will always remain with that SMQ It is recmmended that the use f SMQs is based n the same dictinary as the cded data as mismatches f SMQ and the cded data culd result with unexpected flagged recrds CONCLUSION There are multiple benefits thrugh the use f SMQs and CQs when implemented crrectly. In planning ahead yu can use them t define cnditins early n in cmpund develpment, gruping terms when applicable fr future use. With SMQs, the pre-defined lists are results f an encmpassing effrt s yu can ptentially gain insight t using SMQs that can wrk acrss multiple TAs and be part f a standardized/cnsistent implementatin prcess. In the case where a SMQ may nt cver all medical tpics r safety issues that may be f interest, CQ utilizatin will be f great benefit t yur analysis. As with all things in data standards, an unce f preventin (thughtful cnsideratin) can be wrth a pund f cure (re-wrk). REFERENCES Intrductry Guide fr Standarised MedDRA Queries (SMQs) Versin 19.1, MSSO-DI-6226-19.1.0, 2016, ICH, Available at http://www.meddra.rg/sites/default/files/guidance/file/smq_intguide_19_1_english.pdf List f SMQ Tpics fr Develpment by CIOMS Wrking Grup fr SMQs (as f 1September 2016) Available at http://www.meddra.rg/sites/default/files/page/dcuments/list_f_smq_tpics_fr_website_september_20 16.pdf CDISC ADaM Occurrence Data Structure (OCCDS) (Versin 1.0) Available at www.cdisc.rg CONTACT INFORMATION Yur cmments and questins are valued and encuraged. Cntact the authrs at: Richann Watsn Experis (513) 843-4081 Richann.watsn@experis.cm Karl Miller inventiv Health karl.miller@inventivhealth.cm SAS and all ther SAS Institute Inc. prduct r service names are registered trademarks r trademarks f SAS Institute Inc. in the USA and ther cuntries. indicates USA registratin. Other brand and prduct names are trademarks f their respective cmpanies. 7