Putting Da­ta to Work in a Pan­dem­ic

Me­di­da­ta’s In­tel­li­gent Tri­als is help­ing clin­i­cal tri­als spon­sors dig out of a pan­dem­ic-led slump with pow­er­ful da­ta re­sources and pre­dic­tive analy­sis

Ju­ly marked a turn­ing point in the COVID-19 pan­dem­ic for on­col­o­gy clin­i­cal tri­als. With coro­n­avirus safe­ty pro­to­cols bet­ter es­tab­lished in clin­i­cal set­tings, can­cer pa­tients who had stayed away for months be­gan to re­turn to tri­als. New pa­tients en­ter­ing stud­ies in­creased 20 per­cent ver­sus 2019 base­line lev­els. Just three months ear­li­er, in April, at the be­gin­ning of the pan­dem­ic, new pa­tient re­cruits plunged 40 per­cent ver­sus the pre­vi­ous year.

The Ju­ly turn­around was good news for on­col­o­gy re­searchers, phar­ma com­pa­nies and, of course, can­cer pa­tients. It meant that many on­col­o­gy clin­i­cal tri­al spon­sors could be­gin to re­boot their tri­als and get back to work again on de­sign­ing and test­ing life-sav­ing can­cer drugs. But spon­sors in the on­col­o­gy field, and every oth­er ther­a­peu­tic cat­e­go­ry, still need a lot of help.

Even be­fore the pan­dem­ic hit, clin­i­cal tri­als were be­set by high fail­ure rates due to poor site per­for­mance, re­cruit­ment and da­ta qual­i­ty is­sues. Near­ly 90% of all tri­als were un­able to en­roll pa­tients with­in tar­get time­frames, while each day of de­lay in a drug’s time to mar­ket costs spon­sors an es­ti­mat­ed $600,000 to $8 mil­lion. The COVID-19 pan­dem­ic on­ly com­pound­ed prob­lems for study starts, en­roll­ment, and pa­tient com­ple­tion of vis­its. As the busi­ness be­gins to dig it­self out of a pan­dem­ic-led slump, spon­sors need to know, now more than ever, where to find high per­form­ing sites and how to en­sure that their en­roll­ment and da­ta col­lec­tion ef­forts are on tar­get.

Me­di­da­ta’s In­tel­li­gent Tri­als pro­gram is de­signed to pro­vide an­swers to these ques­tions. The an­a­lyt­ics plat­form aims to im­prove the speed, suc­cess, and qual­i­ty of tri­als through bet­ter study de­sign, en­hanced coun­try and site se­lec­tion, en­roll­ment pre­dic­tions, and re­al-time study track­ing against oth­er sim­i­lar in­dus­try tri­als. Most im­por­tant­ly, it is built on the most com­pre­hen­sive, fine grained and en­riched repos­i­to­ry of clin­i­cal tri­al da­ta in the busi­ness. A long his­to­ry of trust­ed re­la­tion­ships with the in­dus­try’s largest play­ers means Me­di­da­ta has ac­cess not on­ly to un­par­al­leled his­tor­i­cal da­ta re­sources, but al­so ex­ten­sive cur­rent in­sights. Its datasets cov­er 20,000 plus his­tor­i­cal tri­als and 22,000 health­care fa­cil­i­ties with as­so­ci­at­ed in­ves­ti­ga­tors across 94 coun­tries, as well as 6,000 live tri­als hap­pen­ing in re­al time.

The re­al-time da­ta Me­di­da­ta col­lects and cu­rates is es­pe­cial­ly cru­cial in these ex­tra­or­di­nary times. Just as Me­di­da­ta can help its cus­tomers tell which sites did well be­fore the pan­dem­ic, it can see which sites are com­ing back up to speed quick­ly as the pan­dem­ic pro­gress­es. It can help to strat­i­fy the da­ta to look at how pa­tient ac­cru­al dif­fers by coun­try and by ther­a­peu­tic or dis­ease area. And while the pan­dem­ic it­self is un­prece­dent­ed, even now our pre­dic­tive mod­el­ers are look­ing at how we can quan­ti­fy the spe­cif­ic im­pact of the pan­dem­ic on a giv­en site’s op­er­a­tional per­for­mance over time.

Pre­dict­ing the Fu­ture

Me­di­da­ta got a jump start in da­ta an­a­lyt­ics al­most a decade ago. In 2010, the com­pa­ny’s ex­ec­u­tives had the fore­sight to reach out to the biggest names in the phar­ma busi­ness to in­vite them to share da­ta in a co­or­di­nat­ed way. In a spir­it of co­op­er­a­tion, the com­pa­nies agreed to let Me­di­da­ta pool their da­ta to pro­vide analy­sis that would im­prove clin­i­cal tri­als op­er­a­tions for every­one, while pro­tect­ing the da­ta pri­va­cy of in­di­vid­ual pa­tients and re­search or­ga­ni­za­tions. No one else in the in­dus­try has ac­cess to clin­i­cal and clin­i­cal tri­al op­er­a­tional da­ta all in one place from hun­dreds of spon­sors and mul­ti­ple clin­i­cal re­search or­ga­ni­za­tions.

It’s not just the quan­ti­ty of da­ta that makes Me­di­da­ta’s plat­form ex­cep­tion­al. It’s al­so the lev­el of de­tail that it can of­fer. While pub­licly avail­able clin­i­cal tri­al datasets ex­ist, these do not pro­vide the fine grain of in­for­ma­tion that Me­di­da­ta’s datasets do. Its plat­form has in­for­ma­tion from the ac­tu­al ex­e­cu­tion of each study, for ex­am­ple. That da­ta can be com­bined to give deep in­sight in­to per­for­mance of dif­fer­ent dis­eases, coun­tries and sites.
Me­di­da­ta’s da­ta is al­so en­riched through a two-step process that is unique in the busi­ness. Oth­er com­pa­nies use ei­ther al­go­rithms or man­u­al cu­ra­tion to struc­ture their da­ta, but Me­di­da­ta’s ap­proach is a hy­brid of the two. We let the ma­chines do what they do best: learn from the da­ta and ap­ply sys­tems of clas­si­fi­ca­tion. Then we bring sub­ject mat­ter ex­perts in­to the loop to ad­ju­di­cate the find­ings and com­plete struc­tur­ing the da­ta be­fore it is used in analy­sis. We al­so lever­age pub­licly avail­able da­ta from all over the world in­clud­ing tri­al reg­istries and pub­li­ca­tions.

Putting the Da­ta to Work

All of the met­rics and in­sights from the clin­i­cal tri­als da­ta we col­lect can then be fed back in­to the study de­sign process, en­hanc­ing re­cruit­ment and en­roll­ment as well as tim­ing and cost. We use both his­tor­i­cal and re­al time per­for­mance da­ta to train pre­dic­tive mod­els, to an­tic­i­pate fu­ture per­for­mance, to de­ter­mine the like­li­hood that a par­tic­u­lar site will suc­cess­ful­ly en­roll for a par­tic­u­lar kind of tri­al. Study site per­for­mance de­pends on many things, of course, such as how in­ter­est­ing the study drug is and how con­gest­ed the ge­o­graph­i­cal area is in terms of sim­i­lar stud­ies. To­day, it al­so de­pends on COVID-19 rates in a par­tic­u­lar ge­og­ra­phy and the COVID-19 risk fac­tors of the par­tic­u­lar pa­tient pop­u­la­tion that is tar­get­ed.

Our In­tel­li­gent Tri­als pre­dic­tive mod­els can take in­to ac­count hun­dreds of fea­tures, in­clud­ing a tri­al’s coun­try foot­print, as well as the per­for­mance, con­ges­tion, qual­i­ty and ca­pac­i­ty of in­di­vid­ual sites and col­lec­tions of study sites on dif­fer­ent time­lines. Rather than wait­ing for prob­lems to arise, you can gen­er­ate for­ward-look­ing in­sights that ac­cel­er­ate pa­tient en­roll­ment based on his­toric per­for­mance, pre­dic­tive mod­els, in­dus­try trends, and chang­ing con­di­tions in re­al time. And you al­ways have ac­cess to the in­put of our team of ex­perts to help you eval­u­ate these in­sights, in­clud­ing not just da­ta sci­en­tists and tech­nol­o­gists, but ex-reg­u­la­to­ry of­fi­cials and in­ves­ti­ga­tors.

As an ex­am­ple of how this can work, In­tel­li­gent Tri­als re­cent­ly helped a large clin­i­cal tri­als spon­sor ac­cel­er­ate the com­ple­tion of an un­der­per­form­ing phase III tri­al for an ex­per­i­men­tal treat­ment that was fac­ing slowed en­roll­ment. Me­di­da­ta per­formed a study de­sign com­plex­i­ty analy­sis and site-by-site analy­sis to help the spon­sor iden­ti­fy those sites re­quir­ing in­ter­ven­tion and se­lect new sites that had po­ten­tial to speed up time­lines. Ul­ti­mate­ly, the tri­al was ac­cel­er­at­ed by at least 6 months ver­sus an­tic­i­pat­ed time­lines. That helped ad­vance time to mar­ket and re­sult­ed in a huge cost sav­ings for the spon­sor.

Me­di­da­ta has cre­at­ed its In­tel­li­gent Tri­als in­sights to em­pow­er spon­sors to build the best pos­si­ble clin­i­cal tri­als in the most chal­leng­ing of en­vi­ron­ments. Our mis­sion is aligned with theirs: to help bring life sav­ing treat­ments to pa­tients who need them as quick­ly as pos­si­ble, in spite of the pan­dem­ic. We want to help build a bet­ter world.