STATUS REPORTS USING SAS. Dagnlja V. Spuntel1s, Merck Sharp and Dohme Rahway, N.J. faci1~tated. A. System Requirements. 1. A simple input mechanism

Transcription

1 STATUS USING SAS Dagnlja V. Spuntel1s Meck Shap and Dohme Rahway N.J. i: One aspect of phamacutical egulatoy affais is concened with oganizing and audltin clinical poject data which ae to be used in dossies submitted to egulatoy agencies. en subnitting a new dug fo egistatio infomation.must be cpiled on the status of all studies conducted on that poject. Fo each poject clinical studies ae conducted following a pescibed potocol. When comlete they ae summaized domestically in a standad fomat the A to M summay. A to M meely de notes the sections. ABC.D M of the summay. The A to M summay can be futhe classifi@d ag eithe: SOLITARY; having the one study pe potocol INDIVIDUAL; a sepaate summay fo each invetigato following a pat1cula potocol o COMBINED; which includes infoation fom all stgdies done unde the same potocol. All studies including tbose cancelled o fo which thee ae no data in bouse must be accounted fo. The SAS system descibed in this pape was used as a tool in oganizing a lage quantity of such DUlti-sDuce poject data. The aticula poject unde discussion had been undeway fe a numbe of yeats and involved poximately 450 complete o on-going clinical studi@s. This was B. fomidabl amoun of infomation to ogainize much less manipulate by hand-log alone. The assignment equied a knowledge of the histoy s well as the cuent status of the enti9 pdjec. It vas foeseen that many questions would be aised The answe to each c_uestion would euil"e much of the same infomation to be pesented in a slightly diffeent wy. each vaiation equiing mhny hous of epetitive effot. Fo example if the poject ecods wee oganized in study numbe ade then thee would be questions elating to the i.fonat1on in alphabetical ode by investig&tot; if it 'wee an alphabetical listing the. data pobably would be. ne.ede.d by study title o potocol. Keeping half a dozen o moe lists each equiing update as additional infomation a8 eceived did not seem like an optmal solution. A standad epot did not mee.t the equiements eithe. Standad epots can be uite cumbesome to mk voith when only a few p1eces of infomation ae needed f each page of a multipage compute output. The sceening and coelation with othe non-computeized bits of infonation say fom hand tabulations can be tedious time consuming and eo-pone. The situation was complicated by the fact that ove the yeas many depatments and many people had woked on tbe vaious vhases of the poject. Thee was no integated coe of infomation; although the infomation was available only some was computeized. The fi8t step in solvin this poblem 'Waa. to assemble as much elevant ln fomation as possible; the second to decide ho to bst oganize tbe data such that manipulation would be faci1tated. REQUIREMENTS Outlining foeseeable equiements was a stat. Boken down in system and infomation equiements the most impotant of these seemed to be: A. System Requiements B. 1. A simple input mechanism 2. Ability to add addiional study infomation (obsevations) as new study numbes wee assigned 3. No liit on ecod and va1abl length 4. On-line access o infomation 5. Uncomplicated update chaacteistics 6. uncomplicated epotgenetating f@atues 7. Ability to geneate an oveall summay of all ptoject studies 8. Requie at most uinial systems and pogamming suppot Infomation Requiements 1. The data set must include all study numbes. Study numbes ae allocated sequentially by no.iect to signed potocols as they ae eceived. 75

2 2. Identifv studies a. b. c. d. e. which wee cancelled o as91ned a study numbe bu neve state.d which could be duplicates conducted unde paticula potocol on which audits of data accuacy have been completed fo which additional aw data bave'. been eceived since the audit wee used in many instances to deceabe vaiable length and speed input. If a potocol was used fo sveal studies it was m1jch a6ie to indica.te a p-otocol numbe than to include the ntie title (especially in the cae whe One polocol coveed moe than 40 studies). Combined summaies wee also identified by numbe. Sepaate dictionaies identifying the codes wee ceated. These wee set up So that not only peliminay OT woking epots but also detailed -epots could be poduced without unduly inceasing the size of tha basic data set. The vaiables fo this SAS data set ae identified in Table I. SlSEQUENT ADDITIONS AND MODIFIcATIONS TO nata f. g. h. included in a pio doesie fo egistation fo which solitay summaies have been witten. The study and summay ae unique fo which individual summaies have been witten. Applicable to paallel studies done unde the same potocol. but WTitten up individually fo each investigato Once the bu1k of the data had ben ented and numeous epots geneated additonal new epots wee equested fot which all of the equied infomation wag not avai1able in the exiting data set. Rathe than efomatting the existing data set an additional data set was geneated. The new data wee added using as a key the unique identifie of each obsevation - the study numbe. To geneate a epot incopo.ating this new infomation equied only a simple mege. i. included in combined Bummaies. A combined summay incopoates study infomation available fom the many individual studies done unde a paticula potocol..iven the above equiements. SAS was the best choice- of the opions avai1abl@. pimai1y because the whole system could be established and un by someone having a minimal knowledge of computes. DATA lliput Du-ing the couse of the poject. new naming. numbeing and study identification conventions tjee established by the depatment. TI".e new methods wee not howeve etoactive and study infomation eceived peviously emained in the ld fomat. Allowances had to be made fo each convention when assembling thi infomation. Due to the age of the poject the majoity of wok on the individual studies. sti.mla:les etc'. I had aleady been completed. To these popotionally little would have to be added and updated at a late date. This technique was used to add a county code to the data as well as in classifying each study by its unit in the New Dug Application. Company' defined codes wee used instead of county names to speed input and save space. A sepaate county dictionay was set up which included all counties in which clinical - tials might possibly be conducted. REPORT GENERATION Depending on the intended pupose vaious ievels of pogam complexity wee equied to poduce the final epot. In many cases a simple sot and pint was all that was equied to answe the posed question. Output was geneated on-line using a DECWRITER II o outed to the IBM 3800 lase pinte. Examples of epots available unde this system include those aanged and pinted: 1. by potocol in study o investigato ode 2. by county in which the study was conducted The infomation was submitted to batch pocessing via tape. The tape was geneated using the Petec XL-40 key to disk data entty system. The XL-40 was pogammed to pompt fo all vaiables and pe-sceen fo invalid vaiable valus. coections. additions and deletions to this initial SAS data set wee accomplished using the SAS edito pocdu unde TSO. The "standad" obsevation was designed fot a maxdnlim input of l44 chaactes. Codes by type OT Bummay witten B. individual o solitay A-M b. combined A-M to list all pe-maketing studies 76

3 5. to list all studies peviously included in a egistation dossie 6. to list all studies with no data in house 7. to list all cases whee eceipt of new data equied a e-audit of a study B. to show to w-h1ca. unit of tbe New Dug Application a paticula study was assigned " A pic coal summation of the system can be seen in Figue 1. CONCLUSION SAS is vesatile. It can be appjied both to lagescale pojects as well as small It 1s exce11ent fo oganizing data fo vey specific auplications by individuals witb little pogamming backgound. SAS is outstandin in poviding the advantages of computeization to the small scale use. ACKNOWLEDGEMENTS The autho wishes to thank Fank J. Potte fo his techn:fcal assistance. in setting up this system. I: I'. ;. f' f 1" c.!: ;: l " f: ' t "- t I I: i! j; 77

4 TAUE I DATA SET VARIABLES NAME STUDY ST TYPE DEFINITION/PURPOSE study numbe identifies extensions to the study INVEST ALPHI'.BETIC 1nvstigatos name PROTOCOL identifies the study by potocol numbe TITLE identifies the study by an abb.eviated title if the study was not assigned a potocol numbe STATUS identifies studies which ae CANCELLED have O PROTOCOL o ae DUPLICATE AMl (S o I) indicates studies having a SOLItARY {= S) o INn!VIDUAL (= 1) A-M say AM2 and AM3 ALP}IA]ETIC (Cl-991 indicates the paticula= combined summay o summaies in which a study has been included COMBINED ALPHA]ETIC identifies the study pais fo which combined summaies have been witte.n. This was pevalent in the olde studies whee one investigato ould have a diffeent study numbe fo each extension but the summay was witten to include all of the mateial i i'. CR' CEll (CER CER2-9 1 :FMKT o A) identifies the studies having unaudited case epot foms by giving the numbe of foms not yet audited identifies studies included in toe vaious vaious clinical expeince epots! dossies. PMKt identifies pe-maketing studies and A identifie9 mateial that has been audited but not yet included in a elease. COMMENTS gives petinent infomation not coveed elsewbee 78

5 j f h : FIGURE I -; STUDY SUMMARY REGULATORY RHEUU PROTOCOL RAW DATA FILES ;. }. " y f. TAPE SAS DATA SET CORRECTIONS AoolTIDMS PiDC EIITDi ON LINE IBM 3800 BY INVESTIGATOR REPORT ON STUDIES WITH NO DATA BY STUDY 79