As gene ther­a­py man­u­fac­tur­ing con­tin­ues to ma­ture, the fo­cus is shift­ing. Most gene ther­a­py de­vel­op­ers be­gin this work run­ning a sprint, rac­ing to reach first-in-hu­man da­ta as quick­ly as pos­si­ble. But as pro­grams ad­vance to­ward com­mer­cial­iza­tion, the dis­ci­pline re­quired starts to look less like a sprint and more like a marathon. Ear­ly de­vel­op­ment once fo­cused pri­mar­i­ly on demon­strat­ing that a vec­tor could be pro­duced re­pro­ducibly and at suf­fi­cient yield. To­day, as pro­grams ad­vance to­ward piv­otal stud­ies and com­mer­cial­iza­tion, the chal­lenge shifts to a thor­ough, steady jour­ney. Reg­u­la­to­ry agen­cies no longer ask whether a prod­uct can be man­u­fac­tured. They ask whether the process and prod­uct are tru­ly un­der­stood in a way that sup­ports con­sis­ten­cy and safe­ty.

Process char­ac­ter­i­za­tion (PC) pro­vides the sci­en­tif­ic foun­da­tion need­ed to build the da­ta set that en­ables de­vel­op­ers to iden­ti­fy crit­i­cal process pa­ra­me­ters, es­tab­lish ro­bust man­u­fac­tur­ing con­trol strate­gies and demon­strate process un­der­stand­ing as pro­grams ad­vance to­ward com­mer­cial­iza­tion. Like train­ing for a marathon, PC is not about a sin­gle ex­plo­sive ef­fort; it’s the sus­tained, dis­ci­plined work, mile af­ter mile, that builds the en­durance a pro­gram needs to go the dis­tance to com­mer­cial launch. It trans­forms a process that op­er­ates re­li­ably in the hands of its de­vel­op­ers in­to one that can be jus­ti­fied, re­pro­duced, and trust­ed by health au­thor­i­ty re­view­ers. When ex­e­cut­ed well, PC re­duces down­stream risk by strength­en­ing val­i­da­tion strate­gies, sup­port­ing PPQ readi­ness, and en­abling a smoother path to com­mer­cial launch. When treat­ed as both a reg­u­la­to­ry re­quire­ment and a sci­en­tif­ic ex­er­cise, it can be­come a foun­da­tion­al source for in­for­ma­tion re­quests dur­ing BLA or MAA re­view.

With­in the de­vel­op­ment life­cy­cle, process char­ac­ter­i­za­tion serves as the crit­i­cal link be­tween process de­vel­op­ment and com­mer­cial process val­i­da­tion:

It is where the em­pir­i­cal, ex­plorato­ry work of ear­ly de­vel­op­ment is for­mal­ized in­to a de­fen­si­ble con­trol strat­e­gy, ground­ed in sta­tis­ti­cal ev­i­dence rather than in­sti­tu­tion­al knowl­edge. This is pre­cise­ly why PC draws such close reg­u­la­to­ry scruti­ny. Reg­u­la­to­ry bod­ies col­lec­tive­ly ex­pect that crit­i­cal process pa­ra­me­ters (CPPs), proven ac­cept­able ranges (PARs), and the re­sult­ing con­trol strat­e­gy are not as­sert­ed, but demon­strat­ed.

At Forge, our ap­proach to process char­ac­ter­i­za­tion aligns with the Al­liance for Re­gen­er­a­tive Med­i­cine's (ARM) Project A-Gene frame­work, which em­pha­sizes a con­nect­ed sci­en­tif­ic nar­ra­tive rather than a col­lec­tion of in­de­pen­dent stud­ies. Every stage builds up­on the pre­vi­ous one, be­gin­ning with struc­tured risk as­sess­ments, pro­gress­ing through qual­i­fi­ca­tion of rep­re­sen­ta­tive scale-down mod­els, in­to De­sign of Ex­per­i­ments (DoE), and ul­ti­mate­ly sup­port­ing process val­i­da­tion and com­mer­cial readi­ness. To­geth­er, these el­e­ments sup­port reg­u­la­to­ry ex­pec­ta­tions while build­ing the sci­en­tif­ic ba­sis for a ro­bust and scal­able man­u­fac­tur­ing process.

Start­ing char­ac­ter­i­za­tion ear­ly on is what sep­a­rates pro­grams that pace them­selves for the marathon ahead from those left sprint­ing to catch up lat­er.

Be­fore a sin­gle DoE run is ex­e­cut­ed, the ques­tion must be an­swered: what are the great­est risks to prod­uct qual­i­ty? This is the role of the Pa­ra­me­ter Risk As­sess­ment (PRA) and process Fail­ure Mode and Ef­fects Analy­sis (pFMEA). These tools pro­vide a struc­tured frame­work for holis­ti­cal­ly eval­u­at­ing process pa­ra­me­ters in­clud­ing set­points, hold times, and ma­te­r­i­al at­trib­ut­es across each unit op­er­a­tion, and sys­tem­at­i­cal­ly rank­ing their po­ten­tial im­pact on crit­i­cal qual­i­ty at­trib­ut­es (CQAs), which are the mea­sur­able prop­er­ty or char­ac­ter­is­tics that must be con­trolled to en­sure prod­uct qual­i­ty². In gene ther­a­py prod­ucts, rel­e­vant CQAs may in­clude at­trib­ut­es such as pu­ri­ty, po­ten­cy, and in­fec­tiv­i­ty.

The qual­i­ty of a risk as­sess­ment de­pends on the breadth of ex­per­tise in­volved in its de­vel­op­ment. A sci­en­tif­i­cal­ly sound and de­fen­si­ble risk rank­ing re­quires cross-func­tion­al in­put from Process De­vel­op­ment, An­a­lyt­i­cal De­vel­op­ment, Man­u­fac­tur­ing Sci­ence and Tech­nol­o­gy (MSAT), Val­i­da­tion, Qual­i­ty and Reg­u­la­to­ry teams. Each team brings a dis­tinct per­spec­tive to the as­sess­ment of process risk. For de­vel­op­ers work­ing with a CD­MO, this is al­so the mo­ment to bring that part­ner to the ta­ble, as a marathon is best paced with the full team as­sem­bled be­fore race day, not re­cruit­ed mid-course. A pa­ra­me­ter that may ap­pear low risk from a process per­for­mance per­spec­tive may present greater risk when eval­u­at­ed through the lens of an­a­lyt­i­cal ro­bust­ness or clin­i­cal im­pact.

Im­por­tant­ly, risk as­sess­ment is not a one-time mile­stone com­plet­ed ear­ly in de­vel­op­ment. It should be treat­ed as a liv­ing doc­u­ment that evolves along­side the pro­gram, in­cor­po­rat­ing new process knowl­edge, char­ac­ter­i­za­tion da­ta, and man­u­fac­tur­ing ex­pe­ri­ence as the prod­uct ad­vances to­ward com­mer­cial­iza­tion.

Once the right ques­tions are iden­ti­fied, they must be eval­u­at­ed us­ing a mod­el that re­li­ably rep­re­sents com­mer­cial man­u­fac­tur­ing. This is where scale-down mod­el (SDM) qual­i­fi­ca­tion be­comes es­sen­tial.

Think of the SDM as the prac­tice course: it must mir­ror race-day con­di­tions (i.e., com­mer­cial-scale man­u­fac­tur­ing) close­ly enough that a strong train­ing run pre­dicts a strong re­sult on the day that mat­ters.

Reg­u­la­to­ry ex­pec­ta­tions re­quire char­ac­ter­i­za­tion da­ta gen­er­at­ed at lab­o­ra­to­ry or pi­lot scale to be rep­re­sen­ta­tive of com­mer­cial-scale man­u­fac­tur­ing. Com­mon qual­i­fi­ca­tion ap­proach­es in­clude sta­tis­ti­cal equiv­a­lence test­ing (e.g., TOST), com­pa­ra­bil­i­ty as­sess­ments, and mul­ti­vari­ate analy­ses eval­u­at­ing mul­ti­ple qual­i­ty at­trib­ut­es si­mul­ta­ne­ous­ly.

To en­sure SDM qual­i­fi­ca­tion pro­ceeds smooth­ly, sev­er­al con­sid­er­a­tions should be ad­dressed up­front. These in­clude con­firm­ing the avail­abil­i­ty and suit­abil­i­ty of rep­re­sen­ta­tive com­mer­cial-scale batch­es for com­par­i­son, ac­count­ing for ex­pect­ed batch-to-batch vari­abil­i­ty, and defin­ing sci­en­tif­i­cal­ly jus­ti­fied ac­cep­tance cri­te­ria and da­ta in­clu­sion/ex­clu­sion cri­te­ria be­fore ex­e­cu­tion of the qual­i­fi­ca­tion study. The qual­i­fi­ca­tion strat­e­gy should al­so es­tab­lish in ad­vance which process pa­ra­me­ters and CQAs will be com­pared, the sta­tis­ti­cal ap­proach for as­sess­ing scale-de­pen­dent dif­fer­ences, and the ra­tio­nale for de­ter­min­ing equiv­a­lence be­tween the SDM and com­mer­cial-scale process­es. Es­tab­lish­ing these el­e­ments prospec­tive­ly helps dis­tin­guish true scale-de­pen­dent ef­fects from nor­mal process vari­abil­i­ty and pro­vides a clear, de­fen­si­ble frame­work for qual­i­fi­ca­tion. When sup­port­ed by rep­re­sen­ta­tive da­ta and pre­de­fined cri­te­ria, a qual­i­fied SDM can then be con­fi­dent­ly lever­aged for char­ac­ter­i­za­tion of mul­ti­ple unit op­er­a­tions.

With con­fi­dence that the scale-down mod­el is rep­re­sen­ta­tive of man­u­fac­tur­ing, the next chal­lenge is de­sign­ing ex­per­i­ments that gen­er­ate mean­ing­ful process un­der­stand­ing rather than sim­ply con­firm ex­ist­ing as­sump­tions.

This is where DoE pro­vides a crit­i­cal ad­van­tage over one-fac­tor-at-a-time (OFAT) ap­proach­es. While OFAT stud­ies can demon­strate how an in­di­vid­ual pa­ra­me­ter be­haves in iso­la­tion, they are un­able to cap­ture in­ter­ac­tions be­tween pa­ra­me­ters, which are of­ten where the great­est risks to prod­uct qual­i­ty ex­ist. DoE es­tab­lish­es PARs, char­ac­ter­izes pa­ra­me­ter in­ter­ac­tions, and es­tab­lish­es a con­trol strat­e­gy re­flect­ing how the process be­haves across the full op­er­at­ing space.

Ake­so and Sum­mit Ther­a­peu­tics claimed a key win over Keytru­da at the start of Sep­tem­ber, and are now spelling out that sur­vival da­ta, hop­ing it...