The Path to Pro­duc­tiv­i­ty: Tack­ling the Cri­sis in Drug De­vel­op­ment us­ing Im­proved Pre­clin­i­cal Mod­els

“The more pos­i­tive any­one is about the past sev­er­al decades of progress [in phar­ma­ceu­ti­cal de­vel­op­ment], the more neg­a­tive they should be about the strength of coun­ter­vail­ing forces.” These fore­bod­ing words were penned in a sem­i­nal 2012 ar­ti­cle by Jack Scan­nell, au­thor of Eroom’s Law, to il­lu­mi­nate the drug de­vel­op­ment in­dus­try’s pro­duc­tiv­i­ty cri­sis.

Pro­duc­tiv­i­ty mea­sures how ef­fi­cient drug de­vel­op­ment is, of­ten pre­sent­ed as the num­ber of drugs that can be brought to mar­ket giv­en a set amount of ef­fort or in­vest­ment. Con­sid­er phar­ma­ceu­ti­cal de­vel­op­ment in the 1950s: Da­ta pre­sent­ed by Scan­nell and his co-au­thors showed that, with the con­tem­po­rary equiv­a­lent of $1 bil­lion US dol­lars, the in­dus­try was able to pro­duce around 30 new drugs. In con­trast, that same in­vest­ment in 2023 would not even pro­duce one new ther­a­peu­tic (see Fig­ure 1).

Fig­ure 1: Graph show­ing the change R&D ef­fi­cien­cy since 1950.

This de­cline in pro­duc­tiv­i­ty has many in the in­dus­try con­cerned, as low­er pro­duc­tiv­i­ty means steep­er de­vel­op­ment costs and slow­er progress. As a re­sult, drug prices in­crease for pa­tients, who are left des­per­ate­ly wait­ing for ther­a­peu­tic re­lief.

Scan­nell and many oth­ers have worked to di­ag­nose the var­i­ous fac­tors be­hind the cur­rent pro­duc­tiv­i­ty cri­sis—the “coun­ter­vail­ing forces.” Through their ef­forts, many po­ten­tial caus­es have been iden­ti­fied, one of which stands out as pro­found­ly im­pact­ful: im­prov­ing the ac­cu­ra­cy of pre­clin­i­cal mod­els.

Pre­clin­i­cal Drug De­vel­op­ment

Pre­clin­i­cal drug de­vel­op­ment is high­ly spec­u­la­tive: Re­searchers must fore­tell how com­pounds will be­have in the hu­man body and iden­ti­fy the few that are safe and ther­a­peu­tic. To do this, they use mod­els as prox­ies for the hu­man body, and none serve a more promi­nent role than non-hu­man an­i­mal mod­els.

Ro­dents, pri­mates, and oth­er non-hu­man an­i­mal mod­els have long been the gold stan­dard in pre­clin­i­cal tox­i­col­o­gy screen­ing. With com­plex and in­ter­con­nect­ed tis­sues, an­i­mals al­low re­searchers to test the ef­fect of their drug in a dy­nam­ic sys­tem re­sem­bling the hu­man body. As such, an­i­mals are the last fil­ter in the drug de­vel­op­ment process, tasked with weed­ing out tox­ic drugs be­fore clin­i­cal tri­als.

De­spite the ubiq­ui­ty of an­i­mal mod­els, am­ple ev­i­dence in­di­cates they are far from per­fect. Ap­prox­i­mate­ly 90% of drugs en­ter­ing clin­i­cal tri­als fail, with rough­ly 30% of those fail­ures at­trib­uted to un­fore­seen tox­i­c­i­ty. This in­di­cates that, at min­i­mum, an­i­mal mod­els alone are in­suf­fi­cient de­ci­sion-mak­ing tools—too of­ten, they get it wrong, and both pa­tients and drug de­vel­op­ers pay the price.

This fail­ure is cen­tral to the cur­rent pro­duc­tiv­i­ty cri­sis. Though re­searchers now have ac­cess to next-gen­er­a­tion se­quenc­ing, com­bi­na­to­r­i­al chem­istry, and au­toma­tion, drug de­vel­op­ment costs have in­creased near­ly 80-fold since 1950 to a stag­ger­ing $2.3B per ap­proved drug. And ap­prox­i­mate­ly 75% of these costs can be at­trib­uted to fail­ure.

It stands to rea­son that re­duc­ing clin­i­cal tri­al fail­ure rates will im­prove the ef­fi­cien­cy of drug de­vel­op­ment. Not on­ly are failed tri­als ex­pen­sive, but they al­so take up clin­i­cal re­sources that could be used to ad­vance suc­cess­ful drugs. Since clin­i­cal tri­al fail­ure rates re­flect the qual­i­ty of drugs that en­ter tri­als, im­prov­ing this qual­i­ty should in­crease in­dus­try pro­duc­tiv­i­ty.

How should re­searchers go about do­ing this? Re­vis­it­ing his work a decade lat­er, Scan­nell pro­vid­ed pow­er­ful guid­ance: The qual­i­ty of the com­pounds that en­ter clin­i­cal tri­als is a con­se­quence of the pre­clin­i­cal mod­els used to se­lect them, and even small im­prove­ment in the qual­i­ty of the pre­clin­i­cal mod­els—more specif­i­cal­ly, their pre­dic­tive va­lid­i­ty—can have a sub­stan­tial im­pact on pro­duc­tiv­i­ty. En­ter more hu­man-rel­e­vant pre­clin­i­cal mod­els like the Liv­er-Chip.

Im­prov­ing Pro­duc­tiv­i­ty with Or­gan-Chips

In a re­cent­ly pub­lished study, Em­u­late sci­en­tists showed that the Liv­er-Chip—a spe­cial­ized Or­gan-Chip that mim­ics the hu­man liv­er—can iden­ti­fy com­pounds’ po­ten­tial to cause drug-in­duced liv­er in­jury (DILI) far more ac­cu­rate­ly than tra­di­tion­al in vit­ro and an­i­mal mod­els.

Briefly, Or­gan-Chips are three-di­men­sion­al cul­ture sys­tems that com­bine het­ero­ge­neous cell cul­ture, flu­id flow, and sev­er­al fea­tures of the tis­sue mi­croen­vi­ron­ment to mim­ic hu­man or­gan func­tion in an in vit­ro set­ting. Ev­i­dence in­di­cates that hu­man cells cul­tured in Or­gan-Chips be­have re­mark­ably sim­i­lar to their in vi­vo coun­ter­parts. Among many promis­ing ap­pli­ca­tions, these chips are par­tic­u­lar­ly well suit­ed for pre­clin­i­cal tox­i­col­o­gy screen­ing.

In their study, the Em­u­late re­searchers found the Liv­er-Chip to be a high­ly sen­si­tive and spe­cif­ic tool for de­tect­ing he­pa­to­tox­ic com­pounds. In par­tic­u­lar, the Liv­er-Chip showed a sen­si­tiv­i­ty of 87% and speci­fici­ty of 100% against a se­ries of drugs that had pro­gressed in­to the clin­ic af­ter be­ing test­ed in an­i­mal mod­els, on­ly to lat­er be re­vealed as tox­ic when giv­en to pa­tients. There­fore, these drugs well rep­re­sent the cur­rent gap in pre­clin­i­cal tox­i­col­o­gy test­ing, through which some he­pa­to­tox­ic drug can­di­dates evade de­tec­tion and ad­vance in­to clin­i­cal tri­als.

If the Liv­er-Chip can fill the gap left by an­i­mal mod­els, Scan­nell’s frame­work sug­gests that it could pro­found­ly af­fect the in­dus­try’s pro­duc­tiv­i­ty by re­duc­ing the num­ber of safe­ty-re­lat­ed clin­i­cal tri­al fail­ures.

To cal­cu­late how this re­duc­tion may im­pact in­dus­try pro­duc­tiv­i­ty, Em­u­late re­searchers teamed up with Jack Scan­nell to build an eco­nom­ic val­ue mod­el. They showed that ap­ply­ing the Liv­er-Chip in all small-mol­e­cule drug de­vel­op­ment pro­grams could gen­er­ate $3 bil­lion dol­lars an­nu­al­ly for the in­dus­try through im­proved pro­duc­tiv­i­ty. This is ap­prox­i­mate­ly $150M per top phar­ma­ceu­ti­cal com­pa­ny. And, that’s just for the Liv­er-Chip. In ad­di­tion to he­pa­to­tox­i­c­i­ty, car­dio­vas­cu­lar, neu­ro­log­i­cal, im­muno­log­i­cal, and gas­troin­testi­nal tox­i­c­i­ties are among the most com­mon rea­sons clin­i­cal tri­als fail. If Or­gan-Chips can re­duce these clin­i­cal tri­al fail­ures with a sim­i­lar 87% sen­si­tiv­i­ty, the re­sult­ing up­lift in pro­duc­tiv­i­ty could gen­er­ate $24 bil­lion for the in­dus­try an­nu­al­ly—rough­ly $750M to $1B per top phar­ma­ceu­ti­cal com­pa­ny.

Even when ac­count­ing for the cost of in­te­grat­ing and run­ning Liv­er-Chip ex­per­i­ments, the sav­ings from re­duc­ing clin­i­cal tri­al fail­ures are sub­stan­tial. More­over, the ex­tra band­width would per­mit ad­vanc­ing more promis­ing com­pounds. The Em­u­late re­searchers’ work demon­strates that im­prov­ing pro­duc­tiv­i­ty in drug de­vel­op­ment is pos­si­ble, and it starts with de­vel­op­ing bet­ter mod­els. As the in­dus­try em­braces the po­ten­tial of Or­gan-on-a-Chip tech­nol­o­gy and con­tin­ues to ex­plore its ap­pli­ca­tion in var­i­ous ar­eas of drug de­vel­op­ment, there is hope for a fu­ture with im­proved pro­duc­tiv­i­ty and faster de­liv­ery of life-sav­ing ther­a­peu­tics.