How Pa­tients with Epilep­sy Ben­e­fit from Re­al-World Da­ta

Aman­da Shields, Prin­ci­pal Da­ta Sci­en­tist, Sci­en­tif­ic Da­ta Stew­ard

Kei­th Wen­zel, Se­nior Busi­ness Op­er­a­tions Di­rec­tor

Andy Wil­son, Sci­en­tif­ic Lead

Re­al-world da­ta (RWD) has the po­ten­tial to trans­form the drug de­vel­op­ment in­dus­try’s ef­forts to pre­dict and treat seizures for pa­tients with epilep­sy. An­tic­i­pat­ing or con­trol­ling an im­pend­ing seizure can sig­nif­i­cant­ly in­crease qual­i­ty of life for pa­tients with epilep­sy. How­ev­er, be­cause RWD is sec­ondary da­ta orig­i­nal­ly col­lect­ed for oth­er pur­pos­es, the chal­lenge is se­lect­ing, har­mo­niz­ing, and an­a­lyz­ing the da­ta from mul­ti­ple sources in a way that helps sup­port pa­tients.

What is RWD ver­sus RWE?

Ac­cord­ing to the U.S. Food and Drug Ad­min­is­tra­tion (FDA), RWD are “da­ta re­lat­ing to pa­tient health sta­tus or de­liv­ery of health care rou­tine­ly col­lect­ed from a va­ri­ety of sources.¹” Re­al-world ev­i­dence (RWE) is “clin­i­cal ev­i­dence re­gard­ing the us­age and po­ten­tial ben­e­fits or risks of a med­ical prod­uct de­rived from analy­sis of RWD.” Fig­ure 1 il­lus­trates how RWD sup­ports RWE and how that ev­i­dence may be used to sup­port dif­fer­ent seg­ments of the life sci­ences in­dus­try.

Fig­ure 1. FDA Frame­work for RWE

Click on the im­age to see the full-sized ver­sion

RWD is all about con­text

RWD can po­ten­tial­ly help re­searchers an­swer ques­tions quick­ly and ef­fi­cient­ly, and for a broad­er and more rep­re­sen­ta­tive glob­al pa­tient pop­u­la­tion than ran­dom­ized clin­i­cal tri­als (RCTs) alone. By de­sign, RCTs elim­i­nate much of the re­al-world con­text, such as dy­nam­ic treat­ment de­ci­sions, to en­sure da­ta are col­lect­ed in the same way at every site. RCTs are care­ful­ly de­signed to con­trol bias and re­duce any vari­a­tions that could in­flu­ence study re­sults. Pa­tients must meet strict in­clu­sion/ex­clu­sion and ad­her­ence cri­te­ria, so tri­al re­sults may not al­ways be gen­er­al­iz­able to the broad­er pa­tient pop­u­la­tion or re­flect how treat­ments will be used in ways that fit with their lives.

For RWD, the who, what, when, where, why and how da­ta are gen­er­at­ed pro­vide us with im­por­tant con­text. For ex­am­ple, time­stamps on seizure da­ta col­lect­ed from in­tracra­nial EEG read­ings may be im­por­tant da­ta points for dy­nam­i­cal­ly fore­cast­ing seizure events in re­al time. Seizure da­ta col­lect­ed from pa­tient di­aries can pro­vide ev­i­dence of dis­ease pro­gres­sion that may be oth­er­wise dif­fi­cult to cap­ture. RWD and RWE can fur­ther:

  • Ex­pe­dite pa­tient re­cruit­ment and en­roll­ment
  • Ex­pe­dite pa­tient ac­cess to treat­ment
  • Sig­nif­i­cant­ly sup­ple­ment da­ta col­lect­ed from RCTs
  • Bet­ter in­form how to use treat­ment in clin­i­cal prac­tice
  • Per­son­al­ize treat­ment for pa­tients, based on their spe­cif­ic pro­file

How can RWD be used in clin­i­cal stud­ies to im­prove qual­i­ty of life for pa­tients with epilep­sy?

Epilep­sy is a dis­ease that oc­curs in al­most 1% of the pop­u­la­tion world­wide² and is char­ac­ter­ized by re­cur­rent seizures. Al­though many pa­tients can con­trol their dis­ease with reg­u­lar med­ica­tions, about one-third of pa­tients can­not.

By us­ing da­ta to un­der­stand pa­tient-lev­el pat­terns and cy­cles of seizures, spon­sors can com­pare ap­ples to ap­ples more eas­i­ly. “You can’t run all the RCTs you may want to make di­rect com­par­isons be­tween ther­a­pies or com­bi­na­tions of pa­tient pro­files or char­ac­ter­is­tics,” says Aman­da Shields, Prin­ci­pal Da­ta Sci­en­tist, Sci­en­tif­ic Da­ta Stew­ard at Parex­el. “For ex­am­ple, ob­ser­va­tion­al stud­ies in net­work meta-analy­sis (NMA) can be used like a round robin tour­na­ment—to quan­ti­fy in­di­rect com­par­isons.”

Ac­cu­rate­ly pre­dict­ing seizures helps spon­sors de­ter­mine which treat­ments work bet­ter for which pa­tients and pro­vides clin­i­cians tools to de­vise more per­son­al­ized treat­ment plans. Re­mov­ing the un­pre­dictabil­i­ty of and con­trol­ling seizures helps im­prove qual­i­ty of life for pa­tients.

What is the “right” da­ta for epilep­sy stud­ies?

As of this ar­ti­cle’s pub­li­ca­tion, the FDA has not is­sued guid­ance on all as­pects of RWD or RWE. At the end of 2018, the FDA re­leased a draft frame­work for a new Re­al-World Ev­i­dence Pro­gram in which it pro­pos­es to use the RWE Pro­gram to guide gen­er­a­tion of da­ta in sup­port of ap­proval for new in­di­ca­tions or to help sup­port post-ap­proval study re­quire­ments.³ Un­der this frame­work, RWD can be used to sup­ple­ment tra­di­tion­al clin­i­cal tri­al da­ta and im­prove the ef­fi­cien­cy of these tri­als. Ac­cord­ing to the RWE frame­work, one cri­te­ri­on for eval­u­a­tion by the FDA is if the RWD se­lect­ed is fit for use to gen­er­ate da­ta for prod­uct ef­fec­tive­ness de­ci­sions.

At Parex­el, we con­duct a ro­bust fea­si­bil­i­ty process which be­gins with the ques­tion the spon­sor is try­ing to an­swer to en­sure we are col­lect­ing the “right” da­ta for the spe­cif­ic epilep­sy study. Let’s say Spon­sor A wants to eval­u­ate the ef­fec­tive­ness of a drug on pre­vent­ing acute seizures in pa­tients with un­con­trolled or in­com­plete­ly con­trolled seizures on man­age­ment ther­a­py. One im­por­tant source of da­ta will be col­lect­ed from our pa­tient who is wear­ing a non-in­va­sive sen­sor like a smart­watch or oth­er de­vice. Wear­ables can pro­vide sig­nif­i­cant new pre­dic­tive fea­tures. We will com­bine the pre­dic­tive da­ta with his­tor­i­cal da­ta from pa­tient di­aries and elec­tron­ic health records (EHRs) to de­vel­op a com­plete pic­ture.

We then col­lect, har­mo­nize, and an­a­lyze these da­ta, which helps us re­li­ably pre­dict seizures, alert the pa­tient when he may be at risk of an acute event, and prompt him to take the med­ica­tion.

What are the chal­lenges with us­ing RWD for epilep­sy clin­i­cal stud­ies?

There can be chal­lenges with us­ing RWD for any clin­i­cal study. Epilep­sy stud­ies present some unique chal­lenges be­cause the dis­ease is not treat­ed in a sin­gle care set­ting—it might be treat­ed in a hos­pi­tal, pri­ma­ry care, or spe­cial­ist set­ting. Not all EHRs are in­te­grat­ed, so there may be gaps in the da­ta. In the ex­am­ple we men­tioned ear­li­er, our pa­tient may not like his wear­able or may for­get to main­tain his pa­tient di­ary. Da­ta ac­cu­ra­cy can be af­fect­ed if we have ex­ces­sive da­ta gaps due to any or all these is­sues.


  • Lack of pre­scribed reg­u­la­to­ry frame­work
  • Ac­cu­rate, time­ly and use­ful da­ta
  • Link­ing da­ta from mul­ti­ple sources or set­tings
  • Dei­den­ti­fied da­ta
  • Pa­tient con­sent for us­ing their da­ta be­yond a clin­i­cal tri­al set­ting
  • Where da­ta is stored and how to ac­cess it
  • An­a­lyt­ic as­sess­ment and reme­dies for mul­ti­ple sources of bias, e.g., se­lec­tion bias and im­mor­tal time bias
  • Small ver­sus large da­ta sets


RWD, used ap­pro­pri­ate­ly, can help us put to­geth­er the puz­zle pieces that re­veal a very per­son­al sto­ry about epilep­sy pa­tients. Pa­tients can ben­e­fit from the syn­the­sis of all their streams of da­ta, in­clud­ing their own per­spec­tive, an­a­lyt­i­cal­ly in­te­grat­ed in­to both per­son­al sto­ries and ev­i­dence that can speed the de­vel­op­ment of treat­ments that are more pre­cise and per­son­al­ized to the needs of each pa­tient.

¹ U.S. Food and Drug Ad­min­is­tra­tion: Re­al-World Ev­i­dence, 30 Nov 2020,­ence-re­search/sci­ence-and-re­search-spe­cial-top­ics/re­al-world-ev­i­dence
² Baner­jee, P. N., Fil­ip­pi, D. & Allen Hauser, W. The de­scrip­tive epi­demi­ol­o­gy of epilep­sy- a re­view. Epilep­sy Res. 85, 31–45 (2009)
³ PharmEx­ec: Frame­work for FDA’s Re­al-World Ev­i­dence Pro­gram, 31 Jan 2019, https://www.pharmex­­work-fda-s-re­al-world-ev­i­dence-pro­gram-0