For on­col­o­gy drug de­vel­op­ers, re­al-world da­ta (RWD) can be an es­sen­tial tool for in­form­ing de­ci­sions through­out the de­vel­op­ment process. From un­der­stand­ing dis­ease nat­ur­al his­to­ry and defin­ing the pa­tients most like­ly to ben­e­fit from a ther­a­py, to mea­sur­ing and com­par­ing treat­ment re­sponse, RWD helps gauge whether a ther­a­peu­tic hy­poth­e­sis is like­ly to hold be­fore com­mit­ting to a tri­al de­signed to test it. Re­cent­ly, RWD built from se­r­i­al mol­e­c­u­lar resid­ual dis­ease (MRD) test­ing, which tracks cir­cu­lat­ing tu­mor DNA in the blood over time, rep­re­sents a fun­da­men­tal shift in ex­ist­ing da­ta par­a­digms, gen­er­at­ing mul­ti-time point mol­e­c­u­lar and clin­i­cal (EHR cu­rat­ed) in­sights that cap­ture the bi­o­log­i­cal changes un­fold­ing through­out a pa­tien­t's care. Nat­era is pi­o­neer­ing this new da­ta mod­el across its test­ed pa­tient pop­u­la­tion.

Cir­cu­lat­ing tu­mor DNA, or ctD­NA, is ge­net­ic ma­te­r­i­al shed in­to the blood­stream by can­cer cells, and mea­sure­ment of ctD­NA lev­els can be per­formed through a test known as MRD test­ing. Since ctD­NA has a half-life of ap­prox­i­mate­ly one hour, it can pro­vide drug de­vel­op­ers with a re­al-time, quan­ti­ta­tive mea­sure of dis­ease bur­den.¹ Since the test re­quires on­ly a blood sam­ple rather than a scan or biop­sy, it is non-in­va­sive and can be per­formed se­ri­al­ly over the course of treat­ment.*

At the pa­tient lev­el, ctD­NA lev­els ris­ing and falling shows whether a tu­mor is pro­gress­ing or re­spond­ing to treat­ment, guid­ing clin­i­cians and re­searchers to make de­ci­sions such as ad­just­ing treat­ment reg­i­mens or de­ter­min­ing whether a pa­tient is re­spond­ing to an in­ves­ti­ga­tion­al ther­a­py, as ear­ly as weeks af­ter treat­ment ini­ti­a­tion.

Ad­di­tion­al­ly, ul­tra­sen­si­tive as­says, which can de­tect dis­ease at con­cen­tra­tions as low as a few DNA frag­ments in a test tube, can show whether dis­ease still per­sists at a mol­e­c­u­lar lev­el, un­de­tectable by imag­ing.² Af­ter surgery or de­fin­i­tive treat­ment, de­tec­tion of mol­e­c­u­lar resid­ual dis­ease can in­form whether a pa­tient is like­ly to need more treat­ment, months to years be­fore re­lapse be­comes vis­i­ble on a scan.³

Since 2019, Nat­era’s Sig­nat­era^™ MRD test has be­come em­bed­ded in rou­tine on­col­o­gy care: more than half of US on­col­o­gists now or­der it, and tens of thou­sands of pa­tients are test­ed on a month­ly ba­sis.⁴ Since MRD test­ing is re­peat­ed over time, it gen­er­ates a lon­gi­tu­di­nal, mol­e­c­u­lar record of dis­ease bur­den across pa­tients tracked from di­ag­no­sis through treat­ment to out­come.Fig­ure 1. A can­cer pa­tient’s sto­ry told through ctD­NA, a clear­er de­pic­tion of dis­ease ac­tiv­i­ty that cap­tures im­por­tant sig­nals such as treat­ment re­sponse and re­cur­rence in re­al time.

While this lon­gi­tu­di­nal mol­e­c­u­lar record is im­pact­ful for in­di­vid­ual pa­tient care, its val­ue com­pounds when ap­plied across a pop­u­la­tion, ad­dress­ing gaps in ex­ist­ing datasets. Tra­di­tion­al re­al-world da­ta (RWD), typ­i­cal­ly de­rived from elec­tron­ic health records (EHR) and claims records, pro­vides a lon­gi­tu­di­nal clin­i­cal his­to­ry but of­ten lacks a com­plete and gran­u­lar view of tu­mor bi­ol­o­gy, evo­lu­tion, and treat­ment re­sponse or pro­gres­sion. Mul­ti­modal datasets built from re­al-world ge­nomics test­ing can of­fer mol­e­c­u­lar in­sights in­to tu­mor bi­ol­o­gy, but typ­i­cal­ly lack lon­gi­tu­di­nal­i­ty, cap­tur­ing on­ly a sin­gle re­sult per pa­tient rather than the tra­jec­to­ry over time.

Mul­ti-time­point mol­e­c­u­lar RWD com­bines lon­gi­tu­di­nal mol­e­c­u­lar re­sponse sig­nals with lon­gi­tu­di­nal clin­i­cal con­text, pro­duc­ing a more com­plete and gran­u­lar view of dis­ease evo­lu­tion un­der var­i­ous ther­a­peu­tic con­texts and a ba­sis for un­der­stand­ing the un­der­ly­ing dri­vers of re­sponse and re­sis­tance.

Mul­ti-time­point mol­e­c­u­lar da­ta sup­ports two con­nect­ed de­ci­sions in a de­vel­op­ment pro­gram: how to de­sign and de-risk a tri­al be­fore it launch­es, and how to gen­er­ate an ear­ly, re­al-world sig­nal of ther­a­peu­tic ef­fect.

1. De­sign­ing and de-risk­ing tri­als. Re­al-world MRD test­ing and treat­ment pat­terns can re­veal where pa­tients are in­ad­e­quate­ly served by cur­rent stan­dard of care, sur­fac­ing pop­u­la­tions with high un­met need that may ben­e­fit from MRD-guid­ed treat­ment es­ca­la­tion stud­ies. From there, the dataset can al­so mod­el ad­dress­able pop­u­la­tion size, de­fine the op­ti­mal en­roll­ment win­dow, and iden­ti­fy the sites most like­ly to dri­ve en­roll­ment, in­form­ing tri­al de­sign and site strat­e­gy be­fore a sin­gle pa­tient is screened. In cer­tain cas­es, this analy­sis can en­able ear­ly ad­just­ments to im­prove study fea­si­bil­i­ty; for ex­am­ple, ex­pand­ing the pro­posed in­clu­sion/ex­clu­sion cri­te­ria to in­crease like­li­hood of reach­ing the tar­get en­roll­ment.⁵
2. Gen­er­at­ing an ear­ly, re­al-world ef­fi­ca­cy sig­nal. For sin­gle-arm ear­ly-phase stud­ies, de­vel­op­ment teams can build mol­e­c­u­lar-de­fined re­al-world co­horts, matched to a tri­al's tu­mor type, bio­mark­er sta­tus, and treat­ment con­text, and bench­mark re­al-world mol­e­c­u­lar re­sponse rates against in­ves­ti­ga­tion­al da­ta. As mol­e­c­u­lar re­sponse is high­ly cor­re­lat­ed with re­al-world out­comes, this pro­vides an ear­ly sig­nal of ther­a­py ef­fi­ca­cy to in­form down­stream de­vel­op­ment de­ci­sions.⁵

Fig­ure 2. Re­al-world mol­e­c­u­lar re­sponse rates are bench­marked against in­ves­ti­ga­tion­al dataFig­ure 3. Mol­e­c­u­lar re­spon­ders (mR) had longer re­al-world over­all sur­vival (rwOS) vs. mol­e­c­u­lar non-re­spon­ders (mNR)

The val­ue of mul­ti-time­point mol­e­c­u­lar da­ta lies not in its vol­ume, but in how it is con­nect­ed, link­ing lon­gi­tu­di­nal mol­e­c­u­lar sig­nals to the full pa­tient record. The ex­pec­ta­tion is there is enough so­phis­ti­ca­tion to in­form the de­ci­sions that mat­ter most in a de­vel­op­ment pro­gram: which pa­tients to pur­sue, how to de­sign the tri­al that reach­es them, and whether a ther­a­py is work­ing.

Can­cer evolves across the en­tire pa­tient jour­ney, from di­ag­no­sis through treat­ment to out­come. The datasets used to un­der­stand it should be built to re­flect that same evo­lu­tion.

For drug de­vel­op­ers, that shift means the ther­a­pies most like­ly to help can be iden­ti­fied, test­ed, and de­liv­ered to the pa­tients who need them, soon­er.

Ex­plore more here.

*Tu­mor in­formed MRD would re­quire tis­sue up­front for ini­tial as­say de­sign.

Ref­er­ences:

1. Diehl F, Schmidt K, Choti MA, et al. Cir­cu­lat­ing mu­tant DNA to as­sess tu­mor dy­nam­ics. Nat Med. 2008;14(9):985-990. doi:10.1038/nm.1789
2. Rein­ert T, Hen­rik­sen TV, Chris­tensen E, et al. Analy­sis of plas­ma cell-free DNA by ul­tra­deep se­quenc­ing in pa­tients with stages I to III col­orec­tal can­cer. JA­MA On­col. 2019;5(8):1124-1131. doi:10.1001/ja­maon­col.2019.0528
3. Gar­cia-Muril­las I, Ab­bott CW, Cutts RJ, et al. Whole genome se­quenc­ing-pow­ered ctD­NA se­quenc­ing for breast can­cer de­tec­tion. Ann On­col. 2025;36(6):673-681. doi:10.1016/j.an­nonc.2025.01.021
4. Nat­era in­ter­nal da­ta as of Au­gust 2026.
5. Nat­era in­ter­nal analy­sis per­formed on de-iden­ti­fied RWD as of Au­gust 2026.

The "most fa­vored na­tion" drug pric­ing deals that Pfiz­er and Eli Lil­ly signed with the Trump ad­min­is­tra­tion sug­gest the com­pa­nies will give the gov­ern­ment a...