How to use registry data to support FDA decision making: Agency explains in new guidance

Drug­mak­ers look­ing to de­sign a new reg­istry or use an ex­ist­ing one to sup­port a reg­u­la­to­ry de­ci­sion on a drug’s ef­fec­tive­ness or safe­ty will need to con­sult with a new draft guid­ance re­leased Mon­day by the FDA.

The agency’s re­liance on reg­istry da­ta for reg­u­la­to­ry de­ci­sions dates back more than two decades, at least, as in 1998 Bay­er won ap­proval for its an­ti­co­ag­u­lant Re­flu­dan (with­drawn from the mar­ket in 2013 for com­mer­cial rea­sons) based in part on a his­tor­i­cal con­trol group pulled from a reg­istry.

The new 14-page draft guid­ance makes clear that al­though reg­istries may have ad­van­tages over oth­er re­al-world da­ta sources, reg­istries are far from per­fect sources of in­for­ma­tion. FDA had to craft this guid­ance un­der the 21st Cen­tu­ry Cures Act as part of a broad­er frame­work to eval­u­ate the po­ten­tial use of re­al-world ev­i­dence to help sup­port the ap­proval of a new in­di­ca­tion for an ex­ist­ing drug.

In terms of the ad­van­tages of reg­istries, FDA points to their abil­i­ty to col­lect ‘struc­tured and pre­de­ter­mined da­ta el­e­ments’ and ‘of­fer lon­gi­tu­di­nal, cu­rat­ed da­ta about a de­fined pop­u­la­tion of pa­tients and their cor­re­spond­ing dis­ease course, com­pli­ca­tions, and med­ical care. In ad­di­tion, reg­istries can sys­tem­at­i­cal­ly col­lect pa­tient-re­port­ed da­ta that med­ical claims datasets or EHR datasets may lack,’ FDA said.

Ul­ti­mate­ly, FDA said, reg­istry da­ta can be used, when ap­pro­pri­ate, to in­form the de­sign and sup­port ei­ther in­ter­ven­tion­al stud­ies (clin­i­cal tri­als) or non-in­ter­ven­tion­al (ob­ser­va­tion­al) stud­ies.

But reg­istries have lim­i­ta­tions for use in a reg­u­la­to­ry con­text, the agency ex­plains.

For ex­am­ple, ex­ist­ing reg­istries may fo­cus on one dis­ease, with lim­it­ed in­for­ma­tion on co­mor­bid con­di­tions, even af­ter link­age to oth­er da­ta sources. In ad­di­tion, the en­rolled pa­tients may not be rep­re­sen­ta­tive of the tar­get pop­u­la­tion of in­ter­est due to chal­lenges re­lat­ed to pa­tient re­cruit­ment and re­ten­tion. For ex­am­ple, pa­tients with more se­vere dis­ease may be more like­ly to be en­rolled in a reg­istry com­pared to pa­tients with milder dis­ease; or en­rolled pa­tients might have dif­fer­ent self-care prac­tices, so­cioe­co­nom­ic back­grounds, or lev­els of sup­port­ive care ver­sus the en­tire pop­u­la­tion of in­ter­est. These is­sues can po­ten­tial­ly in­tro­duce bias in­to analy­ses.

The agency al­so makes clear that reg­istries are bet­ter suit­ed as a da­ta source for reg­u­la­to­ry pur­pos­es when spon­sors can cap­ture ob­jec­tive end­points, such as death or hos­pi­tal­iza­tion. Where­as sub­jec­tive end­points, such as pain, can be col­lect­ed in a reg­istry, FDA says, but there may be ad­di­tion­al chal­lenges in try­ing to stan­dard­ize such mea­sure­ments.

Fac­tors that FDA con­sid­ers when as­sess­ing the re­li­a­bil­i­ty of reg­istry da­ta (the draft digs in on oth­er da­ta in­tegri­ty ques­tions too) in­clude:

how the da­ta were col­lect­ed (da­ta ac­cru­al)

whether the reg­istry per­son­nel and process­es in place dur­ing da­ta col­lec­tion and analy­sis min­i­mize er­rors and en­sure da­ta in­tegri­ty is suf­fi­cient

whether the reg­istry has pri­va­cy and se­cu­ri­ty con­trols in place to en­sure that the con­fi­den­tial­i­ty and se­cu­ri­ty of da­ta are pre­served.

The draft guid­ance al­so in­cludes a non-ex­haus­tive list of ex­am­ples of da­ta to in­clude in a reg­istry, such as de­mo­graph­ic and clin­i­cal in­for­ma­tion, treat­ment in­for­ma­tion for the dis­ease of in­ter­est, health-re­lat­ed out­come in­for­ma­tion and preg­nan­cy-re­lat­ed in­for­ma­tion.

And FDA ex­plains how spon­sors may con­sid­er link­ing the da­ta on a pa­tient in a reg­istry to the same pa­tient in an­oth­er da­ta sys­tem or sys­tems, such as an­oth­er reg­istry, an EHR, a med­ical claims data­base, or through dig­i­tal health tech­nolo­gies such as sen­sors that al­low for con­tin­u­ous or in­ter­mit­tent record­ing of phys­i­o­log­i­cal and/or be­hav­ioral da­ta (e.g., blood pres­sure, phys­i­cal ac­tiv­i­ty, glu­cose lev­els) or soft­ware ap­pli­ca­tions that are run on gen­er­al-pur­pose com­put­ing plat­forms.

‘Spon­sors should use strate­gies to cor­rect for re­dun­dant da­ta, to re­solve any in­con­sis­ten­cies in the da­ta, and to ad­dress oth­er po­ten­tial prob­lems, such as the abil­i­ty to pro­tect pa­tient pri­va­cy while trans­fer­ring da­ta se­cure­ly,’ the draft adds.

In­dus­try seeks more from FDA on oth­er RWD-fo­cused guid­ance 

Drug­mak­ers and Ama­zon al­so re­cent­ly weighed in on an­oth­er RWE-fo­cused draft guid­ance from FDA, which deals with the use of EHRs and med­ical claims da­ta to aid reg­u­la­to­ry de­ci­sions.

Am­gen, Janssen and Re­gen­eron are seek­ing fur­ther in­fo from FDA on how to meet with and ob­tain time­ly and ef­fi­cient feed­back from the agency on such RWD plans. ‘To en­cour­age ad­vance­ments in RWD use and shared learn­ings be­tween FDA and spon­sors, Janssen en­cour­ages FDA to con­sid­er re­vi­sions to the guid­ance that al­low for a more fea­si­ble ap­proach to a se­ries of fo­cused meet­ings be­tween FDA and spon­sors that al­lows for re­view and feed­back at mile­stones for de­sign­ing, con­duct­ing, and re­port­ing fit-for-pur­pose RWE,’ Na­jat Khan, chief da­ta sci­ence of­fi­cer at Janssen, wrote.

Am­gen is al­so hop­ing to add an ap­pen­dix to the draft guid­ance based on fur­ther col­lab­o­ra­tions with FDA and oth­ers in in­dus­try, in­clud­ing da­ta providers, on a frame­work for da­ta qual­i­ty.

‘We be­lieve that more spe­cif­ic guid­ance is need­ed to en­able da­ta providers to gen­er­ate con­sis­tent as­sess­ment re­ports for da­ta qual­i­ty that will meet the Agency’s needs as part of a spon­sor’s reg­u­la­to­ry sub­mis­sions,’ Am­gen said.

Re­gen­eron said that FDA should make more clear this is a dy­nam­ic top­ic and high­light the agency’s flex­i­bil­i­ty to ex­plore RWD-re­lat­ed con­cerns.

And Ama­zon, which doesn’t fre­quent­ly weigh in on mat­ters re­lat­ed to drug de­vel­op­ment, al­so called on FDA to in­di­cate how spon­sors can sup­ple­ment da­ta us­ing ‘syn­thet­ic in­for­ma­tion, such as from ar­ti­fi­cial in­tel­li­gence (AI)-based tools, nat­ur­al lan­guage pro­cess­ing ser­vices, or dig­i­tal twins (e.g. syn­thet­ic that can serve as the con­trol arm in a study).’

Blair An­der­son, di­rec­tor of AWS Pub­lic Pol­i­cy, al­so sought more gran­u­lar ex­am­ples in the guid­ance:

For ex­am­ple, the dis­cus­sion of miss­ing da­ta con­sid­er­a­tions (page 10) in­cludes il­lus­tra­tive in­for­ma­tion on the ori­gin of omis­sions and how to ad­dress the prob­lem, such as by us­ing a proxy for the miss­ing da­ta. FDA should pro­vide sim­i­lar de­tail in oth­er ar­eas. In par­tic­u­lar, the dis­cus­sion on com­mon da­ta el­e­ments would ben­e­fit from gran­u­lar­i­ty on the da­ta el­e­ments rec­om­mend­ed or that would be re­quest­ed by the agency based on dif­fer­ent sce­nar­ios.
https://endpts.com/how-to-use-registry-data-to-support-fda-decision-making-agency-explains-in-new-guidance/