Comparison between monoclonal antibodies and tyrosine kinase inhibitors

Although mAbs and TKIs both disrupt oncogenic signalling, they differ in target localization, pharmacology, toxicity, resistance mechanisms and clinical positioning. mAbs are large biologics that primarily bind extracellular receptors or ligands, enabling ligand blockade, receptor modulation and, in some cases, immune effector recruitment or payload delivery. In contrast, TKIs are small molecules that enter cells and inhibit intracellular kinase domains, targeting either receptor or non-receptor kinases and, in some cases, multiple signalling pathways (Ref. Reference Smith and Prasad21). The principal molecular, pharmacological and clinical differences between mAbs and TKIs are summarized in Table 2.

Table 2.

Comparative overview of monoclonal antibodies and tyrosine kinase inhibitors

Table 2. Long description

The table consists of four columns: Feature, Monoclonal antibodies, Tyrosine kinase inhibitors, and References.

* Molecular weight: Monoclonal antibodies are approximately 150 k D a (large), while Tyrosine kinase inhibitors are less than 1 k D a (small).

* Target location: Monoclonal antibodies target extracellular receptors or ligands, whereas Tyrosine kinase inhibitors target intracellular kinase domains.

* Administration: Monoclonal antibodies are administered I V or S C; Tyrosine kinase inhibitors are administered orally.

* Half-life: Days to weeks for monoclonal antibodies; hours to days for tyrosine kinase inhibitors.

* B B B penetration: Generally limited for monoclonal antibodies; variable for tyrosine kinase inhibitors, with some achieving C N S levels.

* Immune effector function: Monoclonal antibodies are format dependent (e.g., A D C C or C D C); Tyrosine kinase inhibitors have none.

* Key toxicities: Monoclonal antibodies cause infusion reactions and immune-mediated effects; Tyrosine kinase inhibitors cause skin, G I, hepatic, and cardiovascular effects.

* Resistance mechanisms: Monoclonal antibodies involve antigen loss or bypass pathways; Tyrosine kinase inhibitors involve kinase mutations or amplification.

* Biomarker testing: Often I H C or F I S H (protein-level) for monoclonal antibodies; often N G S (genomic-level) for tyrosine kinase inhibitors.

* Examples: Monoclonal antibodies include Trastuzumab and Bevacizumab; Tyrosine kinase inhibitors include Imatinib and Osimertinib.

These pharmacologic differences also influence clinical use. mAbs are generally administered intravenously or subcutaneously and have prolonged half-lives (Ref. Reference Zhang, Chang, Zhao, Li, Sun, Zhang, Wang and Jin22), whereas TKIs are predominantly oral and their exposure is influenced by metabolism and drug–drug interactions (Ref. Reference Villanueva and Llovet23). Importantly, selected TKIs, including osimertinib, lorlatinib and alectinib, achieve clinically relevant CNS penetration, an advantage in patients with brain metastases (Ref. Reference Malhotra, Mambetsariev, Gilmore, Fricke, Nam, Gallego, Chen, Chen, Amini, Lukas and Salgia24). Toxicities similarly reflect their distinct mechanisms, with mAbs associated with infusion reactions, immune-mediated toxicity or target-specific adverse effects, while TKIs commonly produce kinase-dependent toxicities such as diarrhoea, hepatotoxicity and dermatologic or cardiovascular effects (Ref. Reference Liu and Kurzrock25).

Resistance to TKIs commonly arises through target-site mutations, target amplification, or bypass signalling, whereas mAb resistance may involve antigen loss, impaired immune effector activity, or activation of alternative pathways (Ref. Reference Ramos and Bentires-Alj26). Accordingly, both classes are increasingly incorporated into combination strategies, including dual-target and vertical pathway inhibition. Biomarker selection also differs: mAb therapy is frequently guided by protein-level assessment using IHC or FISH, whereas TKI selection more often relies on genomic alterations identified by molecular profiling (Ref. Reference Sakamoto, Matsubara, Takahama, Yokoyama, Nakamura, Tokito, Okamoto, Akamatsu, Oki, Sato, Tobino, Ikeda, Mori, Mimura, Maeno, Miura, Harada, Nishimura, Hiraoka, Kenmotsu, Fujimoto, Shimokawa, Yamamoto and Nakagawa27). Within NCCN-guided precision oncology, TKIs are particularly prominent in molecularly defined, oncogene-addicted tumours, whereas mAbs have broader applications, including combination with chemotherapy, ICIs and ADCs. Notably, mAbs and TKIs can also be combined to achieve complementary inhibition of the same or interconnected signalling pathways. Thus, mAbs and TKIs are complementary rather than competing platforms, providing extracellular and intracellular therapeutic strategies that can be integrated according to tumour biology, biomarker profile and clinical context.

Most clinically relevant targeted agents in NCCN guidelines

Immune checkpoint inhibitors

The PD-1/PD-L1 axis is a key immune checkpoint exploited by tumours to evade immune surveillance. PD-1 inhibitors (e.g., nivolumab, pembrolizumab, dostarlimab) act on T cells, while PD-L1 inhibitors (e.g., atezolizumab, durvalumab) target tumour-expressed ligands to disrupt inhibitory signalling (Refs Reference Lin, Kang, Chen, Zeng, Li, Xiong, Yi and Xiang28, Reference Liu, Li and Li29). Figure 2 presents a schematic representation of the mechanism of action of ICIs.

Figure 2.

ICI mechanism of action. Left (priming): Dendritic cell (DC) presents tumour antigens to naïve T cells (MHC-TCR) with CD28-B7 costimulation. CTLA-4 is blocked by ipilimumab to enhance activation. Right (effector): Activated T cells recognize tumour cells. PD1/PDL1 inhibition by pembrolizumab, nivolumab or durvalumab restores cytotoxic antitumor activity. CTLA-4: cytotoxic T-lymphocyte-associated protein 4; DC: dendritic cell; MHC: major histocompatibility complex; PD-1: programmed cell death protein 1; PD-L1: programmed cell death ligand 1.

Figure 2. Long description

The diagram is divided into two main sections by a vertical dashed line.

Left Section: Priming phase in the Lymph node.

At the top, a purple D C interacts with a blue T cell. An arrow labeled Activation and expansion points to a cluster of T cells. Below, a magnified view shows the interface between a Dendritic cell and a Naive T cell.

- M H C on the Dendritic cell binds to T C R on the T cell, labeled Activation signals.

- B 7 on the Dendritic cell binds to C D 2 8 on the T cell.

- Below these, the drug Ipilimumab, a Y-shaped antibody, blocks C T L A-4 on the T cell, preventing it from binding to B 7. This is labeled T cell inhibitory signals blocked.

Right Section: Effector phase in Peripheral tissue.

At the top, an activated blue T cell interacts with a red Cancer cell. Below, a magnified view shows the interface between an Activated T cell and a Tumor cell.

- T C R on the T cell binds to M H C on the Tumor cell, labeled Immune attack with a rightward arrow.

- Below this, P D-1 on the T cell is shown being blocked by antibodies Pembrolizumab or Nivolumab.

- Simultaneously, P D-L 1 forward slash 2 on the Tumor cell is blocked by the antibody Durvalumab.

- These interventions are collectively labeled T cell inhibitory signals blocked, restoring the T cell’s ability to attack the tumor.

PD-1 inhibitors: pembrolizumab and nivolumab across solid tumours

Pembrolizumab and nivolumab are the most extensively incorporated PD-1 inhibitors in NCCN guidelines, reflecting their central role in contemporary oncology. Pembrolizumab is listed in over 100 NCCN indications, either as monotherapy or in rational combinations, while nivolumab has nearly 100 recommendations, frequently in dual immune checkpoint blockade regimens with ipilimumab or relatlimab, underscoring the growing clinical relevance of combination immunotherapy strategies (Ref. Reference Liu, Xu, Cheng, Lin, Jiang, Yu, Zhang, Lu and Zhang30). Beyond numerical representation, their positioning across neoadjuvant, adjuvant, metastatic and maintenance settings highlights a structural shift in cancer care toward immune-based backbone regimens.

A major inflection point in precision oncology occurred in 2017, when pembrolizumab became the first tumour-agnostic therapy approved for MSI-H/dMMR solid tumours, based on the KEYNOTE-158 and KEYNOTE-164 trials (Refs 31–Reference Le, Kim, Van Cutsem, Geva, Jäger, Hara, Burge, O’Neil, Kavan, Yoshino, Guimbaud, Taniguchi, Elez, Al-Batran, Boland, Crocenzi, Atreya, Cui, Dai, Marinello, Diaz and André33). This milestone established a mutation-driven, tissue-independent treatment paradigm that reshaped guideline development and accelerated the integration of comprehensive genomic profiling into routine practice. Subsequent expansions into tumour mutational burden–high disease and perioperative settings, including early-stage NSCLC and high-risk triple-negative breast cancer, further illustrate the migration of PD-1 inhibition into curative-intent strategies (Refs Reference Wakelee, Liberman, Kato, Tsuboi, Lee, Gao, Chen, Dooms, Majem, Eigendorff, Martinengo, Bylicki, Rodríguez-Abreu, Chaft, Novello, Yang, Keller, Samkari and Spicer34, Reference Schmid, Cortes, Pusztai, McArthur, Kümmel, Bergh, Denkert, Park, Hui, Harbeck, Takahashi, Foukakis, Fasching, Cardoso, Untch, Jia, Karantza, Zhao, Aktan, Dent and O’Shaughnessy35).

Together, pembrolizumab and nivolumab illustrate the transition of ICIs from late-line salvage therapies to foundational components of first-line and early-stage treatment algorithms, as reflected by their broad incorporation into NCCN guidelines. Across gastrointestinal malignancies, distinct patterns of use have emerged that reflect differences in biomarker dependence and clinical breadth. In oesophageal and gastro-oesophageal junction cancers, nivolumab has a clearly defined adjuvant role following neoadjuvant chemoradiotherapy and surgery (CheckMate-577) (Ref. Reference Kelly, Ajani, Kuzdzal, Zander, Van Cutsem, Piessen, Mendez, Feliciano, Motoyama, Lièvre, Uronis, Elimova, Grootscholten, Geboes, Zafar, Snow, Ko, Feeney, Schenker, Kocon, Zhang, Zhu, Lei, Singh, Kondo, Cleary and Moehler36), whereas pembrolizumab is preferentially used in biomarker-enriched populations, including MSI-H/dMMR tumours and PD-L1–high disease (KEYNOTE-590) (Ref. Reference Sun, Shen, Shah, Enzinger, Adenis, Doi, Kojima, Metges, Li, Kim, Cho, Mansoor, Li, Sunpaweravong, Maqueda, Goekkurt, Hara, Antunes, Fountzilas, Tsuji, Oliden, Liu, Shah, Bhagia and Kato37). Importantly, nivolumab-based regimens demonstrated survival benefits irrespective of PD-L1 expression in CheckMate-648 (Ref. Reference Kato, Doki, Chau, Xu, Wyrwicz, Motoyama, Ogata, Kawakami, Hsu, Adenis, El Hajbi, Di Bartolomeo, Braghiroli, Holtved, Makino, Blum Murphy, Amaya-Chanaga, Patel, Hu, Matsumura, Kitagawa and Ajani38), suggesting broader applicability. In gastric cancer, both pembrolizumab- and nivolumab-based combinations have demonstrated benefit in broad trial populations, although the magnitude of benefit increases with higher PD-L1 expression and is particularly marked in MSI-H/dMMR tumours (Refs Reference Rha, Oh, Yañez, Bai, Ryu, Lee, Rivera, Alves, Garrido, Shiu, Fernández, Li, Lowery, Çil, Cruz, Qin, Luo, Pan, Wainberg, Yin, Bordia, Bhagia and Wyrwicz39–Reference Janjigian, Shitara, Moehler, Garrido, Salman, Shen, Wyrwicz, Yamaguchi, Skoczylas, Campos Bragagnoli, Liu, Schenker, Yanez, Tehfe, Kowalyszyn, Karamouzis, Bruges, Zander, Pazo-Cid, Hitre, Feeney, Cleary, Poulart, Cullen, Lei, Xiao, Kondo, Li and Ajani42). The distinct clinicopathological and molecular characteristics of MSI-H gastric cancer have also been examined (Ref. [Reference Zhang, Huang, Li, Yan, Xie, Wang and Zhao43]). Accordingly, PD-L1 cutoffs and the strength of guideline recommendations vary according to regimen, tumour location and regulatory context.

In colorectal cancer, both agents have transformed the management of MSI-H/dMMR metastatic disease. Pembrolizumab significantly improved progression-free and overall survival (OS) compared with chemotherapy in KEYNOTE-177 (Ref. Reference Diaz, Shiu, Kim, Jensen, Jensen, Punt, Smith, Garcia-Carbonero, Benavides, Gibbs, de la Fourchardiere, Rivera, Elez, Le, Yoshino, Zhong, Fogelman, Marinello and Andre44), while nivolumab, alone or with ipilimumab, produced durable responses in CheckMate-142 (Ref. Reference André, Lonardi, Wong, Lenz, Gelsomino, Aglietta, Morse, Van Cutsem, McDermott, Hill, Sawyer, Hendlisz, Neyns, Abdullaev, Memaj, Lei, Dixon, Kopetz and Overman45). However, ICIs remain largely ineffective in microsatellite-stable colorectal cancer, where response rates are consistently low (Ref. Reference Guven, Kavgaci, Erul, Syed, Magge, Saeed, Yalcin and Sahin46), reinforcing the necessity of molecular stratification and limiting broader applicability.

In hepatocellular carcinoma, pembrolizumab and nivolumab were initially introduced in the post-sorafenib setting, based on meaningful response rates and durability (KEYNOTE-224, CheckMate-040) (Refs Reference Kudo, Finn, Edeline, Cattan, Ogasawara, Palmer, Verslype, Zagonel, Fartoux, Vogel, Sarker, Verset, Chan, Knox, Daniele, Yau, Gurary, Siegel, Wang, Cheng and Zhu47, Reference Yau, Kang, Kim, El-Khoueiry, Santoro, Sangro, Melero, Kudo, Hou, Matilla, Tovoli, Knox, Ruth He, El-Rayes, Acosta-Rivera, Lim, Neely, Shen, Wisniewski, Anderson and Hsu48). Although subsequent phase three trials did not meet predefined statistical thresholds (Ref. Reference Finn, Ryoo, Merle, Kudo, Bouattour, Lim, Breder, Edeline, Chao, Ogasawara, Yau, Garrido, Chan, Knox, Daniele, Ebbinghaus, Chen, Siegel, Zhu and Cheng49), combination approaches incorporating dual checkpoint blockade or anti-angiogenic therapy have since demonstrated superior outcomes and reshaped first-line management (Refs Reference Llovet, Kudo, Merle, Meyer, Qin, Ikeda, Xu, Edeline, Ryoo, Ren, Masi, Kwiatkowski, Lim, Kim, Breder, Kumada, Cheng, Galle, Kaneko, Wang, Mody, Dutcus, Dubrovsky, Siegel and Finn50–Reference Ren, Xu, Bai, Xu, Cang, Du, Li, Lu, Chen, Guo, Chen, Liu, Jia, Wu, Wang, Shao, Zhang, Shan, Meng, Wu, Gu, Yang, Liu, Shi, Gao, Yin, Cui, Huang, Xing, Mao, Teng, Qin, Wang, Xia, Yin, Yang, Chen, Wang, Zhou and Fan52). In this context, the role of PD-1 inhibitors is defined more by clinical factors than by validated predictive biomarkers.

Renal cell carcinoma illustrates a distinct paradigm in which pembrolizumab- and nivolumab-based strategies dominate across disease stages. Pembrolizumab is uniquely approved as adjuvant monotherapy for high-risk clear cell RCC (KEYNOTE-564) (Ref. Reference Powles, Tomczak, Park, Venugopal, Ferguson, Symeonides, Hajek, Gurney, Chang, Lee, Sarwar, Thiery-Vuillemin, Gross-Goupil, Mahave, Haas, Sawrycki, Burgents, Xu, Imai, Quinn and Choueiri53) and is frequently combined with VEGF-TKIs in the metastatic setting (Refs Reference Powles, Plimack, Soulières, Waddell, Stus, Gafanov, Nosov, Pouliot, Melichar, Vynnychenko, Azevedo, Borchiellini, McDermott, Bedke, Tamada, Yin, Chen, Molife, Atkins and Rini54, Reference Motzer, Alekseev, Rha, Porta, Eto, Powles, Grünwald, Hutson, Kopyltsov, Méndez-Vidal, Kozlov, Alyasova, Hong, Kapoor, Alonso Gordoa, Merchan, Winquist, Maroto, Goh, Kim, Gurney, Patel, Peer, Procopio, Takagi, Melichar, Rolland, De Giorgi, Wong, Bedke, Schmidinger, Dutcus, Smith, Dutta, Mody, Perini, Xing and Choueiri55). Nivolumab, in contrast, retains an important role as monotherapy in IO-naïve patients (CheckMate-025) (Ref. Reference Motzer, Escudier, George, Hammers, Srinivas, Tykodi, Sosman, Plimack, Procopio, McDermott, Castellano, Choueiri, Donskov, Gurney, Oudard, Richardet, Peltola, Alva, Carducci, Wagstaff, Chevreau, Fukasawa, Tomita, Gauler, Kollmannsberger, Schutz, Larkin, Cella, McHenry, Saggi and Tannir56) and as part of dual ICI therapy with ipilimumab, which offers durable benefit in intermediate- and poor-risk disease (CheckMate-214) (Ref. Reference Tannir, Albigès, McDermott, Burotto, Choueiri, Hammers, Barthélémy, Plimack, Porta, George, Donskov, Atkins, Gurney, Kollmannsberger, Grimm, Barrios, Tomita, Castellano, Grünwald, Rini, Jiang, Desilva, Fedorov, Lee and Motzer57). These patterns reflect a balance between depth and durability of response versus the need for rapid cytoreduction.

In NSCLC and melanoma, PD-1 inhibitors have redefined both early- and advanced-stage treatment. Perioperative nivolumab and pembrolizumab significantly improve pathological response and event-free survival in resectable NSCLC (Refs Reference Wakelee, Liberman, Kato, Tsuboi, Lee, Gao, Chen, Dooms, Majem, Eigendorff, Martinengo, Bylicki, Rodríguez-Abreu, Chaft, Novello, Yang, Keller, Samkari and Spicer34, Reference Forde, Spicer, Lu, Provencio, Mitsudomi, Awad, Felip, Broderick, Brahmer, Swanson, Kerr, Wang, Ciuleanu, Saylors, Tanaka, Ito, Chen, Liberman, Vokes, Taube, Dorange, Cai, Fiore, Jarkowski, Balli, Sausen, Pandya, Calvet and Girard58, Reference Cascone, Awad, Spicer, He, Lu, Sepesi, Tanaka, Taube, Cornelissen, Havel, Karaseva, Kuzdzal, Petruzelka, Wu, Pujol, Ito, Ciuleanu, de Oliveira Muniz Koch, Janssens, Alexandru, Bohnet, Moiseyenko, Gao, Watanabe, Coronado Erdmann, Sathyanarayana, Meadows-Shropshire, Blum and Provencio Pulla59), while both agents improve survival in metastatic disease across PD-L1 expression levels when combined with chemotherapy or other ICIs (Refs Reference Hellmann, Paz-Ares, Bernabe Caro, Zurawski, Kim, Carcereny Costa, Park, Alexandru, Lupinacci, de la Mora Jimenez, Sakai, Albert, Vergnenegre, Peters, Syrigos, Barlesi, Reck, Borghaei, Brahmer, O’Byrne, Geese, Bhagavatheeswaran, Rabindran, Kasinathan, Nathan and Ramalingam60, Reference Paz-Ares, Luft, Vicente, Tafreshi, Gümüş, Mazières, Hermes, Çay Şenler, Csőszi, Fülöp, Rodríguez-Cid, Wilson, Sugawara, Kato, Lee, Cheng, Novello, Halmos, Li, Lubiniecki, Piperdi and Kowalski61). In melanoma, pembrolizumab and nivolumab are considered clinically equivalent as first-line monotherapy (Ref. Reference Schachter, Ribas, Long, Arance, Grob, Mortier, Daud, Carlino, McNeil, Lotem, Larkin, Lorigan, Neyns, Blank, Petrella, Hamid, Zhou, Ebbinghaus, Ibrahim and Robert62), whereas nivolumab-based dual checkpoint blockade offers superior efficacy at the expense of increased toxicity (Ref. Reference Hodi, Chiarion-Sileni, Gonzalez, Grob, Rutkowski, Cowey, Lao, Schadendorf, Wagstaff, Dummer, Ferrucci, Smylie, Hill, Hogg, Marquez-Rodas, Jiang, Rizzo, Larkin and Wolchok63). Emerging perioperative strategies further support earlier incorporation of PD-1 inhibition in high-risk disease (Refs Reference Patel, Othus, Chen, Wright, Yost, Hyngstrom, Hu-Lieskovan, Lao, Fecher, Truong, Eisenstein, Chandra, Sosman, Kendra, Wu, Devoe, Deutsch, Hegde, Khalil, Mangla, Reese, Ross, Poklepovic, Phan, Onitilo, Yasar, Powers, Doolittle, In, Kokot, Gibney, Atkins, Shaheen, Warneke, Ikeguchi, Najera, Chmielowski, Crompton, Floyd, Hsueh, Margolin, Chow, Grossmann, Dietrich, Prieto, Lowe, Buchbinder, Kirkwood, Korde, Moon, Sharon, Sondak and Ribas64, Reference Blank, Lucas, Scolyer, van de Wiel, Menzies, Lopez-Yurda, Hoeijmakers, Saw, Lijnsvelt, Maher, Pulleman, Gonzalez, Torres Acosta, van Houdt, Lo, Kuijpers, Spillane, Klop, Pennington, Zuur, Shannon, Seinstra, Rawson, Haanen, Ch’ng, Naipal, Stretch, van Thienen, Rtshiladze, Wilgenhof, Kapoor, Meerveld-Eggink, Grijpink-Ongering, van Akkooi, Reijers, Gyorki, Grünhagen, Speetjens, Vliek, Placzke, Spain, Stassen, Amini-Adle, Lebbé, Faries, Robert, Ascierto, van Rijn, van den Berkmortel, Piersma, van der Westhuizen, Vreugdenhil, Aarts, Stevense-den Boer, Atkinson, Khattak, Andrews, van den Eertwegh, Boers-Sonderen, Hospers, Carlino, de Groot, Kapiteijn, Suijkerbuijk, Rutkowski, Sandhu, van der Veldt and Long65).

Thus, pembrolizumab and nivolumab exemplify how ICIs have become structurally embedded across NCCN guidelines. Pembrolizumab is more frequently positioned in biomarker-defined or perioperative contexts, while nivolumab demonstrates broader applicability through combination strategies and PD-L1–independent activity. Together, they illustrate the maturation of immunotherapy into a cornerstone of modern, guideline-driven oncology.

From a guideline-structural perspective, the widespread incorporation of pembrolizumab and nivolumab also reflects a broader evolution in evidence generation and regulatory alignment. Accelerated approvals based on surrogate endpoints such as pathological complete response, objective response rate (ORR), or event-free survival have increasingly informed NCCN recommendations, particularly in perioperative and high-risk settings. This trend has enabled earlier clinical adoption of PD-1 inhibitors while long-term survival data continue to mature. At the same time, the differential toxicity profiles and dosing schedules of pembrolizumab and nivolumab influence real-world treatment selection, especially in elderly patients or those with comorbidities. As novel biomarkers, immune signatures and combination strategies continue to emerge, PD-1 inhibitors are likely to undergo further refinement rather than expansion, with future guideline updates emphasizing optimal sequencing, de-escalation strategies and patient selection to maximize durable benefit while minimizing unnecessary immune-related toxicity.

Other ICIs: PD-1, PD-L1 and CTLA-4-directed agents

Dostarlimab, a PD-1-targeting mAb within the same therapeutic class as pembrolizumab and nivolumab, is distinguished by its highly selective, biomarker-driven clinical indications. Approved across 16 NCCN indications, its most impactful use is in first-line treatment of primary or recurrent dMMR/MSI-H endometrial cancer (EC) and as neoadjuvant therapy for non-resectable dMMR/MSI-H CRC. The GARNET trial demonstrated an ORR of 45.5% in EC with a favourable immune-related toxicity profile (Ref. Reference André, Berton, Curigliano, Sabatier, Tinker, Oaknin, Ellard, de Braud, Arkenau, Trigo, Gravina, Kristeleit, Moreno, Abdeddaim, Vano, Samouëlian, Miller, Boni, Torres, Gilbert, Brown, Dewal, Dabrowski, Antony, Zografos, Veneris and Banerjee66), while the phase three RUBY trial demonstrated significant PFS and OS benefits in combination with carboplatin-paclitaxel (Ref. Reference Powell, Bjørge, Willmott, Novák, Black, Gilbert, Sharma, Valabrega, Landrum, Gropp-Meier, Stuckey, Boere, Gold, Segev, Gill, Gennigens, Sebastianelli, Shahin, Pothuri, Monk, Buscema, Coleman, Slomovitz, Ring, Herzog, Balas, Grimshaw, Stevens, Lai, McCourt and Mirza67).

A small study in dMMR locally advanced rectal cancer reported 100% clinical complete responses with neoadjuvant dostarlimab, suggesting a paradigm shift toward non-operative, organ-preserving approaches (Ref. Reference Cercek, Lumish, Sinopoli, Weiss, Shia, Lamendola-Essel, El Dika, Segal, Shcherba, Sugarman, Stadler, Yaeger, Smith, Rousseau, Argiles, Patel, Desai, Saltz, Widmar, Iyer, Zhang, Gianino, Crane, Romesser, Pappou, Paty, Garcia-Aguilar, Gonen, Gollub, Weiser, Schalper and Diaz68). Dostarlimab’s clinical impact lies in precise, biomarker-guided deployment that maximizes efficacy while limiting overtreatment in selected populations.

Durvalumab, a PD-L1 inhibitor with 14 NCCN guideline indications, offers a mechanistically distinct approach by targeting the ligand rather than the receptor. Its major impact has been in NSCLC, where the PACIFIC trial established consolidation durvalumab after concurrent chemoradiotherapy, and the AEGEAN trial supported perioperative durvalumab in resectable stage II–III NSCLC (Refs Reference Spigel, Faivre-Finn, Gray, Vicente, Planchard, Paz-Ares, Vansteenkiste, Garassino, Hui, Quantin, Rimner, Wu, Özgüroğlu, Lee, Kato, de Wit, Kurata, Reck, Cho, Senan, Naidoo, Mann, Newton, Thiyagarajah and Antonia69, Reference Heymach, Harpole, Mitsudomi, Taube, Galffy, Hochmair, Winder, Zukov, Garbaos, Gao, Kuroda, Ostoros, Tran, You, Lee, Antonuzzo, Papai-Szekely, Akamatsu, Biswas, Spira, Crawford, Le, Aperghis, Doherty, Mann, Fouad and Reck70). Benefit is not uniform across molecular subgroups; exploratory PACIFIC data did not demonstrate a clear advantage in EGFR-mutated tumours, for which targeted consolidation strategies have subsequently emerged. Thus, clinical and molecular context remains important when selecting post-chemoradiotherapy treatment.

CTLA-4, an inhibitory receptor on T cells, restricts early immune activation by competing with CD28 for B7 ligands on antigen-presenting cells (Ref. Reference Sobhani, Tardiel-Cyril, Davtyan, Generali, Roudi and Li71). Ipilimumab was the first ICI to demonstrate an OS benefit in solid tumours, transforming metastatic melanoma therapy. In the MDX010-20 trial, it produced durable survival despite a modest ORR (Ref. Reference McDermott, Haanen, Chen, Lorigan and O’Day72). Although largely replaced by PD-1/PD-L1 inhibitors as first-line monotherapy, ipilimumab remains clinically relevant in combination regimens, particularly with nivolumab, inducing deep and durable responses. Mechanistically, CTLA-4 blockade enhances naïve T-cell priming and reduces regulatory T-cell suppression, albeit with higher immune-related toxicity.

Non-ICI monoclonal antibodies

Non-ICI mAbs remain foundational in oncology, acting through receptor or ligand blockade (HER2, EGFR), angiogenesis inhibition (VEGF/VEGFR) and immune-mediated cytotoxicity via ADCC or CDC (CD20). Increasingly, they function as engineered platforms, including ADCs and bispecific Abs. These agents complement ICIs by directly inhibiting oncogenic signalling, modulating tumour vasculature or selectively delivering cytotoxic payloads. Advances in ADC design and bispecific Abs such as amivantamab, targeting dual pathways, have demonstrated efficacy, particularly in lung cancer. Non-ICI mAbs are widely incorporated into NCCN guidelines across solid and haematological malignancies, from metastatic to perioperative and maintenance settings, shaping modern combination strategies (Refs Reference Zhang, Chen, Zhou, Liang, Zhang, Su, Zhang and Qu73, Reference Tsuchikama, Anami, Ha and Yamazaki74). Figure 3 illustrates the key mechanisms involved.

Figure 3.

Illustration of molecular targets of approved therapies on a tumour cell. Drugs targeting cell surface receptors (e.g., trastuzumab against HER2, cetuximab against EGFR) and intracellular signalling pathways (e.g., vemurafenib for BRAF^V600E, trametinib for MEK, bortezomib blocking the proteasome, venetoclax antagonizing BCL-2) are depicted. ADC: antibody–drug conjugate.

Figure 3. Long description

A diagram of a tumor cell membrane and cytoplasm.

At the top, extracellular drugs target membrane receptors from left to right.

* Brentuximab targets C D 30, leading to internalization of the A D C and lysosomal degradation, releasing M M A E which causes microtubule disruption.

* Rituximab targets C D 20.

* Trastuzumab targets H E R 2.

* Larotrectinib and Entrectinib target T R K A/B/C.

* Entrectinib and Repotrectinib target R O S 1.

* Crizotinib targets A L K, M E T, and R O S 1.

* Cabozantinib targets M E T and V E G F R 2.

* Selpercatinib, Sunitinib, Pazopanib, and Lenvatinib target R E T and P D G F R.

* Bevacizumab binds V E G F, blocking its interaction with V E G F R to inhibit angiogenesis.

* Cetuximab and Panitumumab target E G F R.

Intracellularly, signaling cascades flow downward.

* On the left, a pathway moves from P I 3 K to A K T to m T O R C 1. Everolimus and Sirolimus inhibit m T O R C 1, which otherwise leads to 4 E B P 1 and S 6 K 1 for cell growth and survival.

* On the right, a pathway moves from S O S to R A S to R A F to M E K to M A P K, leading to cell proliferation and differentiation. Dabrafenib and Vemurafenib inhibit R A F. Trametinib inhibits M E K.

A central box details other mechanisms.

* Lenalidomide acts on the E 3 Ubiquitin Ligase Complex, leading to degradation of I K F Z 1/3 transcription factors, causing apoptosis of myeloma cells and T/N K cell activation.

* Bortezomib causes proteasome inhibition and N F-kappa B pathway disruption.

* Venetoclax acts as a B C L-2 inhibitor.

Trastuzumab transformed HER2-positive breast cancer therapy and is used across first-line, adjuvant, neoadjuvant and subsequent-line settings, either alone or with pertuzumab or tucatinib. Landmark trials (NSABP B-31 (Ref. Reference Perez, Romond, Suman, Jeong, Sledge, Geyer, Martino, Rastogi, Gralow, Swain, Winer, Colon-Otero, Davidson, Mamounas, Zujewski and Wolmark75), HERA (Ref. Reference Cameron, Piccart-Gebhart, Gelber, Procter, Goldhirsch, de Azambuja, Castro, Untch, Smith, Gianni, Baselga, Al-Sakaff, Lauer, McFadden, Leyland-Jones, Bell, Dowsett and Jackisch76)) demonstrated significant improvements in DFS and OS, establishing trastuzumab as the standard of care across breast, upper gastrointestinal and selected ECs, marking a milestone in precision oncology.

Bevacizumab, a VEGF-A inhibitor, was the first approved anti-angiogenic mAb, with >40 indications, including NSCLC, mCRC, HCC and ovarian cancer. It is particularly transformative as maintenance therapy in advanced epithelial ovarian cancer. ICON7 and GOG-0218 demonstrated significant PFS improvements when bevacizumab was combined with chemotherapy and continued as maintenance in advanced ovarian cancer, especially in high-risk patients (Ref. Reference Tewari, Burger, Enserro, Norquist, Swisher, Brady, Bookman, Fleming, Huang, Homesley, Fowler, Greer, Boente, Liang, Ye, Bais, Randall, Chan, Ferriss, Coleman, Aghajanian, Herzog, DiSaia, Copeland, Mannel, Birrer and Monk77).

Rituximab targets CD20 and is foundational in hematologic oncology. The GELA trial demonstrated that adding rituximab to CHOP substantially increased complete-response rates and prolonged event-free survival and OS in diffuse large B-cell lymphoma (Ref. Reference Coiffier, Lepage, Briere, Herbrecht, Tilly, Bouabdallah, Morel, Van Den Neste, Salles, Gaulard, Reyes, Lederlin and Gisselbrecht78). R-CHOP remains a standard regimen more than two decades later. Rituximab therefore improved the curative potential of CHOP-based therapy rather than converting an otherwise uniformly palliative disease into a curable one.

EGFR inhibitors cetuximab and panitumumab are transformative in RAS wild-type mCRC, introducing biomarker-driven negative selection. Activity is minimal in unselected patients, while BRAF status and tumour sidedness further refine patient selection. Recommended in first- or subsequent-line therapy, these agents improve response rates and PFS, occasionally enabling conversion to resectability (Ref. Reference Douillard, Siena, Cassidy, Tabernero, Burkes, Barugel, Humblet, Bodoky, Cunningham, Jassem, Rivera, Kocákova, Ruff, Błasińska-Morawiec, Šmakal, Canon, Rother, Oliner, Wolf and Gansert79).

Brentuximab vedotin, a CD30-directed ADC linked to MMAE, allows selective cytotoxic delivery. Initially approved for relapsed/refractory cHL post-autologous transplant, brentuximab showed high response rates and durable remissions (Refs Reference Fabbri, Cencini, Gozzetti, Schiattone and Bocchia80–Reference Straus, Długosz-Danecka, Connors, Alekseev, Illés, Picardi, Lech-Maranda, Feldman, Smolewski, Savage, Bartlett, Walewski, Ramchandren, Zinzani, Hutchings, Munoz, Lee, Kim, Advani, Ansell, Younes, Gallamini, Liu, Little, Fenton, Fanale and Radford82). ECHELON-1 and ECHELON-2 expanded its role in frontline therapy for CD30+ malignancies, particularly in older or transplant-ineligible patients, highlighting the therapeutic potential of ADC platforms (Ref. Reference Horwitz, O’Connor, Pro, Trümper, Iyer, Advani, Bartlett, Christensen, Morschhauser, Domingo-Domenech, Rossi, Kim, Feldman, Menne, Belada, Illés, Tobinai, Tsukasaki, Yeh, Shustov, Hüttmann, Savage, Yuen, Zinzani, Miao, Bunn, Fenton, Fanale, Puhlmann and Illidge83).

Predictive biomarkers in targeted therapy

The clinical utility of mAbs and TKIs is inseparable from the biomarkers that guide their selection. Predictive biomarkers can be categorized by molecular level and testing methodology, each with distinct clinical applications and limitations.

Genomic biomarkers are the most clinically established. Single-gene alterations including EGFR mutations, ALK/ROS1/RET/NTRK fusions, BRAF V600E, MET exon 14 skipping, KRAS G12C, FGFR alterations and BRCA1/2 mutations serve as companion diagnostic targets for corresponding therapies. HER2 amplification has expanded from breast cancer to gastric, colorectal, biliary and ECs (Ref. Reference Adesoye, Dumbrava, Raghav, Sahin, Chen, Lee, Javle, Pant, Alhalabi, Le, Valero, Pohlmann and Meric-Bernstam113), now encompassing HER2-low categories defining ADC eligibility. Composite genomic biomarkers, including MSI-H/dMMR and TMB-H, provide broader predictive signals for ICI responsiveness and underpin most tumour-agnostic approvals. PD-L1 expression remains widely used for ICI selection, but its predictive value varies across tumour types, scoring systems and antibody clones, and its standalone utility is increasingly questioned (Ref. Reference Vainer, Badve, Rajadurai, Soares, Viale, Büttner, Nuti, Juco, Krishnan and Tsao114).

Liquid biopsy biomarkers, particularly circulating tumour DNA (ctDNA), complement tissue-based profiling and are used as companion diagnostics for several targeted therapies (Ref. Reference Lockwood, Borsu, Cankovic, Earle, Gocke, Hameed, Jordan, Lopategui, Pullambhatla, Reuther, Rumilla, Tafe, Temple-Smolkin, Terraf and Tsimberidou115). Beyond genotyping, ctDNA dynamics can provide early information on treatment response and emerging resistance. In curative settings, postoperative ctDNA is a strong prognostic marker for recurrence. In the DYNAMIC trial, ctDNA-guided management reduced adjuvant chemotherapy use while maintaining similar 5-year recurrence-free and OS (Ref. Reference Tie, Wang, Lo, Lahouel, Cohen, Wong, Shapiro, Harris, Khattak, Burge, Lee, Harris, McLachlan, Horvath, Karapetis, Shannon, Singh, Yip, Ananda, Underhill, Ptak, Silliman, Dobbyn, Popoli, Papadopoulos, Tomasetti, Kinzler, Vogelstein and Gibbs116). The GOZILA study further associated ctDNA-guided matching with improved outcomes in advanced gastrointestinal tumours (Ref. Reference Nakamura, Ozaki, Ueno, Komatsu, Yuki, Esaki, Taniguchi, Sunakawa, Yamaguchi, Kato, Denda, Nishina, Takahashi, Satoh, Yasui, Satake, Oki, Kato, Ohta, Matsuhashi, Goto, Okano, Ohtsubo, Yamazaki, Yamashita, Iida, Yuasa, Bando and Yoshino117). Important limitations include reduced sensitivity in low-tumour-shedding settings and false-positive findings caused by clonal haematopoiesis (Ref. Reference Hu, Ulrich, Supplee, Kuang, Lizotte, Feeney, Guibert, Awad, Wong, Jänne, Paweletz and Oxnard118).

Transcriptomic and proteomic biomarkers remain largely investigational for mAb or TKI selection. Gene-expression signatures such as Oncotype DX and MammaPrint have established prognostic and predictive roles in breast cancer but limited application to targeted-therapy selection (Ref. Reference Piccart, van’t Veer, Poncet, Lopes Cardozo, Delaloge, Pierga, Vuylsteke, Brain, Vrijaldenhoven, Neijenhuis, Causeret, Smilde, Viale, Glas, Delorenzi, Sotiriou, Rubio, Kümmel, Zoppoli, Thompson, Matos, Zaman, Hilbers, Fumagalli, Ravdin, Knox, Tryfonidis, Peric, Meulemans, Bogaerts, Cardoso and Rutgers119). Proteomic and phosphoproteomic analyses may provide functional information beyond genomic drivers, although broadly validated clinical-grade assays remain limited (Ref. Reference Boulos, Yousof Idres and Efferth120). Reverse-phase protein arrays (RPPA) enables high-throughput, pathway-level profiling of tumour samples (Ref. Reference Akbani, Becker, Carragher, Goldstein, de Koning, Korf, Liotta, Mills, Nishizuka, Pawlak, Petricoin, Pollard, Serrels and Zhu121). Decision-support programs such as MD Anderson’s Precision Oncology Decision Support (PODS) framework integrate genomic findings with curated clinical evidence to prioritize investigational biomarker-matched therapy (Ref. Reference Meric-Blernstam, Johnson, Holla, Bailey, Brusco, Chen, Routbort, Patel, Zeng, Kopetz, Davies, Piha-Paul, Hong, Eterovic, Tsimberidou, Broaddus, Bernstam, Shaw, Mendelsohn and Mills122), while Clinical Proteomic Tumour Analysis Consortium (CPTAC) resources are expanding proteogenomic reference data across tumour types (Ref. Reference Rodriguez, Zenklusen, Staudt, Doroshow and Lowy123).

Integrated multi-omic approaches are also beginning to reveal targetable metabolic vulnerabilities that extend beyond canonical oncogenic drivers. Using combined metabolomic and transcriptomic profiling, Zheng et al. demonstrated that gamabufotalin exerts anti-HCC activity by downregulating STAMBPL1 and modulating amino acid metabolism, illustrating how multi-omic integration can link metabolic reprogramming to candidate therapeutic targets (Ref. Reference Zheng, Xu, Cai, Zhu, Xiao, Peng and Chen124).

The evolution from single-gene companion diagnostics to multigene NGS panels, now recommended as preferred initial testing in NSCLC and increasingly adopted across tumour types, and to liquid biopsy-based profiling has fundamentally expanded the biomarker toolkit. As actionable biomarkers continue to multiply, equitable access to comprehensive molecular profiling, harmonized testing standards and clinical decision-support tools will be critical to translating complex biomarker data into treatment decisions.

Resistance to targeted therapy

Resistance to mAbs and TKIs may be primary (de novo, present at treatment initiation) or acquired (developing after an initial response), and both substantially limit the durability of targeted therapy across tumour types. Acquired resistance mechanisms fall into three broad categories: on-target reactivation of the original signalling pathway, most classically through secondary kinase domain mutations that reduce drug binding affinity (e.g., EGFR T790M/C797S, BCR-ABL kinase domain mutations); activation of bypass signalling pathways that circumvent the inhibited target (e.g., MET amplification, HER2 amplification or AXL upregulation in EGFR-mutant NSCLC); and pathway-indifferent mechanisms, including histologic transformation (e.g., to small-cell lung cancer) and epithelial-mesenchymal transition (Ref. Reference Lovly and Shaw125). Analogous mechanisms occur with mAbs, including antigen loss or downregulation, impaired Fc-mediated immune effector function and activation of compensatory signalling (Ref. Reference Smith126).

Clonal evolution and intratumoral heterogeneity are now recognized as central drivers of this process. Rare pre-existing resistant subclones can be selected for and expanded under the selective pressure of targeted therapy, producing minimal residual disease that ultimately re-emerges as the dominant, drug-resistant population (Ref. Reference Lim and Ma127). Multiregion and longitudinal single-cell sequencing have revealed that this branching evolution is shaped by both genetic and non-genetic (epigenetic, transcriptional plasticity, drug-tolerant persister states) mechanisms (Ref. Reference McDonald and Dedhar128), and that the tumour microenvironment actively participates in shaping resistant phenotypes through hypoxia, stromal crosstalk and immune editing (Ref. Reference Khalaf, Hana, Chou, Singh, Mackiewicz and Kaczmarek129).

Acquired resistance increasingly appears to extend beyond secondary mutations and bypass signalling to encompass metabolic and post-translational adaptations. Zheng et al. reported that DAPK2 dysfunction drives Mic60 lactylation and mitochondrial cristae remodelling, activating mitochondrial metabolism and promoting EGFR-TKI resistance and metastasis in lung cancer, reinforcing the concept that epigenetic and metabolic plasticity are integral components of the resistant phenotype (Ref. Reference Zheng, She, Han, Tang, Dou, Lu, Huang, Lin, Wu, He, Du, Wu, Zhang, Hu, Zhu, Wang, Huang, Wu and He130).

Several strategies are used clinically to delay or overcome resistance. Sequential use of next-generation inhibitors designed to retain activity against known resistance mutations (e.g., osimertinib for T790M, third- and fourth-generation ALK inhibitors) remains the most established approach (Ref. Reference Cooper, Sequist and Lin131). Combination strategies, including vertical pathway inhibition (e.g., BRAF plus MEK), dual-target blockade (e.g., trastuzumab plus pertuzumab) and TKI–ICI combinations, can delay bypass signalling, although efficacy and toxicity are regimen-specific. Serial ctDNA analysis can identify emerging resistance alterations, sometimes before radiographic progression; however, the clinical benefit of changing treatment solely on the basis of molecular progression remains under investigation (Ref. Reference Oxnard, Paweletz, Kuang, Mach, O’Connell, Messineo, Luke, Butaney, Kirschmeier, Jackman and Jänne132). Single-cell and spatial multi-omic approaches are also being explored to characterize resistant subclones and drug-tolerant persister states (Ref. Reference Kim, Lee, Hong and Park133).

Next-generation targeted platforms

Beyond conventional mAbs, engineered antibody formats are rapidly expanding the therapeutic reach of antibody-based oncology. Approximately 400 ADCs and over 200 bispecific antibodies are currently in clinical development globally. ADCs couple a tumour-targeting antibody, a chemical linker and a cytotoxic payload, enabling selective intracellular delivery of highly potent agents such as microtubule inhibitors and topoisomerase I inhibitors. Beyond conventional single-payload designs, next-generation ADC architectures include bispecific and biparatopic ADCs (addressing antigen heterogeneity and resistance), dual-payload ADCs, conditionally active (probody) ADCs that enhance tumour specificity and immune-stimulating antibody conjugates that combine cytotoxic and immunomodulatory activity (Ref. Reference Alradwan, Alnefaie, Al Fayez, Aodah, Majrashi, Alturki, Fallatah, Almughem, Tawfik and Alshehri134).

Bispecific antibodies simultaneously engage two distinct antigens or epitopes, either bridging cytotoxic immune cells to tumour cells (T-cell engagers such as epcoritamab and glofitamab in relapsed/refractory large B-cell lymphoma) or blocking two oncogenic pathways concurrently, as with amivantamab (EGFR/MET) in EGFR exon 20 insertion-mutant NSCLC. Bispecific ADCs (bsADCs), which combine dual-antigen recognition with cytotoxic payload delivery, represent an emerging hybrid modality designed to improve internalization and address antigen heterogeneity, although none has yet achieved regulatory approval (Ref. Reference Klein, Brinkmann, Reichert and Kontermann135).

Antibody fragments and nanobodies (single-domain, camelid-derived VHH antibodies, ~15 kDa) may address limitations of full-size mAbs, including restricted tumour penetration, uneven biodistribution and prolonged systemic exposure. Their small size, stability and engineered flexibility support multivalent and multispecific constructs, including cell engagers, radionuclide conjugates and nanobody-based CAR-T/CAR-NK platforms. Nanobody-based products have reached regulatory approval, although oncology-specific applications remain limited, and many candidates targeting HER2, EGFR, PD-L1, VEGF and CTLA-4 remain in preclinical or early clinical development (Ref. Reference Li, Zhou, Cao, Hu, Wang, He, Jiang, Li, Chen, Gong, Luo, Deng and Luo136).

These next-generation platforms extend antibody-based precision oncology beyond conventional mAbs. The clinical impact of advanced ADC design is illustrated by trastuzumab deruxtecan, which improved outcomes relative to trastuzumab emtansine in HER2-positive metastatic breast cancer (Ref. Reference Cortés, Kim, Chung, Im, Park, Hegg, Kim, Tseng, Petry, Chung, Iwata, Hamilton, Curigliano, Xu, Huang, Kim, Chiu, Pedrini, Lee, Liu, Cathcart, Bako, Verma and Hurvitz137).

Beyond antibody-based delivery, rational prodrug and nanoassembly engineering offers a complementary strategy for tumour-selective cytotoxic delivery. Feng et al. showed that precise molecular tailoring of doxorubicin prodrugs, varying fatty-alcohol chain length across a disulfide linker, optimizes the balance between nanoassembly stability and tumour-responsive drug release, prolonging circulation, enhancing tumour accumulation and improving antitumor efficacy, underscoring how molecular-level design can enhance the therapeutic index of payloads delivered through non-antibody platforms (Ref. Reference Feng, Wang, Xu, Zheng, Zhou, Wang and Luo138).

Safety and toxicity management

Safety profiles differ substantially among non-ICI mAbs, ICIs, TKIs and ADCs. Non-ICI mAbs may cause infusion reactions and target-dependent toxicities, including left-ventricular dysfunction with HER2-directed therapy, acneiform rash with EGFR-directed antibodies, hypertension or proteinuria with VEGF inhibition, and hepatitis B reactivation with anti-CD20 therapy. ICIs can produce immune-mediated multiorgan toxicity through loss of self-tolerance (Ref. Reference Petrelli, Dottorini, Rossitto, Parati, Ghilardi, Borgonovo, Dognini, Petrò, D’Alessio, Colombo, Viglino, Nicolì, Faggiano, Tarantini, Callea and Ghidini139). TKI adverse effects may be on-target, off-target, or idiosyncratic and are often managed through treatment interruption, dose modification, supportive care, or switching agents (Refs Reference Shyam Sunder, Sharma and Pokharel140, Reference Roskoski141). ADC toxicity reflects the combined influence of target expression, antibody, linker, payload and tissue distribution; interstitial lung disease is a major concern for selected ADCs (Ref. Reference Petrelli, Dottorini, D’Alessio, Tarantini, Bonetti, Rossitto, Cherri, Zaniboni and Ghidini142). NCCN therefore uses drug-class-specific toxicity guidance rather than a single management algorithm (Ref. Reference Thompson, Schneider, Brahmer, Zaid, Achufusi, Armand, Berkenstock, Bermas, Braaten, Budde, Chokshi, Crees, Davies, Deng, Gesthalter, Jain, Jain, Jallouk, Kaffenberger, Khalil, Lechner, Li, Marr, McGettigan, McPherson, Medina, Mohindra, Olszanski, Oluwole, Patel, Prosek, Reddy, Reid, Ryan, Ryder, Salman, Santomasso, Shofer, Sosman, Wang, Zaha, Zucker, Lyons, Awotiwon and Hang143).

Table 3 summarizes selected class-specific toxicities and management principles for ICIs, TKIs and ADCs. Non-ICI mAb toxicities are described above because they vary substantially according to the therapeutic target.

Table 3.

Selected toxicities and management principles for immune checkpoint inhibitors, tyrosine kinase inhibitors and antibody-drug conjugates

Table 3. Long description

The table consists of five columns: Toxicity, I C Is (mechanism/management), T K Is (mechanism/management), A D Cs (mechanism/management), and References.

* Pulmonary: I C Is involve immune-mediated pneumonitis treated with steroids or discontinuation. T K Is show less common drug-specific I L D. A D Cs show drug-specific I L D or pneumonitis requiring treatment interruption and corticosteroids.

* Cardiac: I C Is involve rare myocarditis treated with high-dose steroids and potentially abatacept or ruxolitinib. T K Is show reversible L V E F decline, hypertension, or Q T prolongation. A D Cs may cause L V E F decline requiring periodic monitoring.

* Hepatic: I C Is involve immune-mediated hepatitis treated with steroids. T K Is show direct hepatotoxicity requiring dose modification. A D Cs show payload-dependent hepatotoxicity requiring liver function monitoring.

* Dermatologic: I C Is involve immune-mediated dermatitis treated with steroids. T K Is show acneiform rash or hand-foot syndrome. A D Cs show payload-dependent skin toxicity requiring supportive care.

* Management paradigm: I C Is focus on immunosuppression and grade-based rechallenge. T K Is focus on dose reduction or sequencing. A D Cs focus on baseline risk assessment and prompt interruption for suspected I L D.

Viewed this way, the recurring theme across drug classes is not the toxicity itself but the shift in monitoring strategy required as targeted agents move earlier into the disease course. When ICIs and ADCs were confined to later-line short-duration use, toxicity management was reactive; as these agents move into adjuvant, maintenance and multi-year treatment settings, proactive baseline risk stratification and scheduled surveillance become necessary to sustain long-term exposure safely, a principle equally applicable to resistance-driven sequencing of next-generation TKIs, which can carry distinct off-target toxicity profiles relative to their predecessors.

Real-world evidence and disparities in access

Real-world data consistently demonstrate that the survival benefits of targeted therapies and ICIs observed in clinical trials do not translate uniformly across diverse patient populations. Multiple registry-based studies indicate that real-world outcomes frequently fall short of pivotal trial benchmarks, particularly among trial-ineligible patients with poor performance status, organ dysfunction or brain metastases (Refs Reference Garitaonaindia, Luczak, Ruhlmann, Petersen, Friis, Borch, Guldbrandt, Svane, Schmidt, Bastholt, Ellebaek and Donia144–Reference Gan, Stukalin, Meyers, Dudani, Grosjean, Dolter, Ewanchuk, Goutam, Sander, Wells, Pabani, Cheng, Monzon, Morris, Basappa, Pal, Wood, Donskov, Choueiri and Heng146). Although some real-world cohorts achieve outcomes comparable to those reported in clinical trials (Ref. Reference Wang, Hubbard and Mamtani147), treatment of trial-ineligible populations may be associated with early toxicity or treatment discontinuation without providing a corresponding OS benefit (Ref. Reference Parikh, Min, Wileyto, Riaz, Gross, Cohen, Hubbard, Long and Mamtani148). Importantly, current guidelines emphasize that advanced age alone should not preclude targeted treatment, supported by favourable real-world outcomes reported for agents such as osimertinib in elderly patients (Ref. Reference Yamamoto, Asahina, Honjo, Sumi, Nakamura, Ito, Kikuchi, Hommura, Honda, Yokoo, Fujita, Oizumi, Morita, Ikezawa, Tanaka, Kimura, Sasaki, Sukoh, Takashina, Harada, Dosaka-Akita and Isobe149). An additional advantage of targeted therapies is that, in selected biomarker-defined populations, treatment decisions may be less dependent on ECOG performance status than those for conventional cytotoxic chemotherapy, particularly when the targeted agent has a favourable tolerability profile. This may allow treatment of patients who would otherwise be considered unsuitable for intensive chemotherapy. Nevertheless, the broader efficacy gap highlights the limitations of extrapolating highly selected clinical trial populations to the heterogeneous and clinically complex populations encountered in routine oncology practice.

Translating trial efficacy into routine practice is further complicated by challenges related to treatment adherence and toxicity management. Because most TKIs are administered orally, adherence represents an important determinant of treatment effectiveness, with nonadherence associated with inferior outcomes in malignancies such as CML and NSCLC (Refs Reference Cheng, Cui, Li, Wang and Li150, Reference Mangues-Bafalluy, Bernardez, Martínez-Sesmero, Navarro-Ruiz, Martín-Conde, Rubio-Salvador, Rius-Perera, Gilabert-Sotoca, Domínguez López and Callejo Mellén151). This may be driven by high pill burden, financial constraints, treatment-related adverse effects and limited health literacy (Ref. Reference Skrabal Ross, Gunn, Suppiah, Patterson and Olver152). Compounding these challenges, toxicity profiles observed in routine practice may be more heterogeneous than those reported in controlled clinical trials, where eligibility criteria, monitoring intensity and standardized management algorithms can differ substantially from routine care. irAEs may require prolonged immunosuppressive treatment, while early TKI intolerance can lead to dose reductions, treatment interruptions or premature discontinuation, potentially compromising therapeutic benefit.

Underlying these challenges are persistent inequities in access to precision oncology. Racial and ethnic minorities, older adults and rural populations remain underrepresented in many foundational oncology trials, limiting the generalizability of evidence across diverse populations (Ref. Reference Zhao, Jiang, Shao and Tang153). These disparities extend into clinical practice, where minority and low-income populations may have reduced access to molecular diagnostic testing and comprehensive genomic profiling, potentially limiting access to biomarker-matched therapies (Ref. Reference Hage Chehade, Ozay, Jo, Gebrael, Sayegh, Ostrowski, Srivastava, Garg, Hardy, Murdock, Narang, Mathew Thomas, Li, Maughan, Roy, Agarwal and Swami154). Furthermore, a substantial proportion of patients with metastatic cancer do not receive guideline-concordant systemic therapy (Ref. Reference Fox, Ward, Alexander, Osarogiagbon, Smith, Studts and Silvestri155) and disparities in OS may persist even when effective treatments are available (Ref. Reference Shah, Lei, Thakur, Buradkar, Alharbi, Ombada, Yao, Chua, Yu, Badve, Khoury, Madabhushi, Baid, Kalinsky, Kabraji and Gandhi156). Recognizing this translational gap, contemporary oncology guidelines increasingly emphasize consideration of geographic, socioeconomic and structural barriers when applying treatment recommendations.

Addressing these challenges will require not only continued development of effective targeted therapies and biomarkers but also interventions aimed at improving molecular testing, treatment adherence, toxicity management and equitable access to precision oncology.

Cross-guideline perspectives: NCCN, ESMO and ASCO

While NCCN, ESMO and ASCO are broadly concordant on the core direction of modern oncology (biomarker-driven targeted therapy, immunotherapy integration and multidisciplinary care), they differ most in methodology and how benefit is graded. NCCN provides consensus-driven, continuously updated treatment algorithms but offers weaker comparative and value-assessment tooling (Ref. Reference Pluchino and D’Amico157); ASCO applies rigorous systematic-review-based methodology but updates less frequently (Ref. Reference Loblaw, Prestrud, Somerfield, Oliver, Brouwers, Nam, Lyman and Basch158); ESMO uniquely formalizes both target actionability (ESCAT) and magnitude of benefit (ESMO-MCBS) (Refs Reference Mateo, Chakravarty, Dienstmann, Jezdic, Gonzalez-Perez, Lopez-Bigas, Ng, Bedard, Tortora, Douillard, Van Allen, Schultz, Swanton, André and Pusztai159, Reference Cherny, Oosting, Dafni, Latino, Galotti, Zygoura, Dimopoulou, Amaral, Barriuso, Calles, Kiesewetter, Gomez-Roca, Gyawali, Piccart, Passaro, Roitberg, Tarazona, Trapani, Curigliano, Wester, Zarkavelis, Zielinski and de Vries160).

These philosophical differences have measurable consequences: when ESCAT and ESMO-MCBS were applied retrospectively to NCCN-recommended genome-targeted therapies, only 12% of supporting trials reached substantial benefit (grade 4–5), with most relying on single-arm phase I/II trials using overall response rate rather than survival, raising questions about the evidentiary strength behind some ‘preferred’ NCCN designations (Ref. Reference Tibau, Hwang, Avorn and Kesselheim161).

Convergence is nonetheless the norm where evidence is mature: societies align on the core actionable NSCLC alterations and on multigene NGS as the preferred strategy, and international consensus (JSCO-ESMO-ASCO-JSMO-TOS) has been reached for tumour-agnostic MSI-H/dMMR and NTRK-fusion indications (Ref. Reference Yoshino, Pentheroudakis, Mishima, Overman, Yeh, Baba, Naito, Calvo, Saxena, Chen, Takeda, Cervantes, Taniguchi, Yoshida, Kodera, Kitagawa, Tabernero, Burris and Douillard162).

Differences persist where evidence is evolving or resource considerations differ: ESMO restricts routine broad NGS to defined tumour types, citing cost and low incremental yield (Ref. Reference Mosele, Westphalen, Stenzinger, Barlesi, Bayle, Bièche, Bonastre, Castro, Dienstmann, Krämer, Czarnecka, Meric-Bernstam, Michiels, Miller, Normanno, Reis-Filho, Remon, Robson, Rouleau, Scarpa, Serrano, Mateo and André163), whereas NCCN endorses comprehensive profiling more liberally. Also, NCCN, ESMO and CSCO diverge on selected TKI sequencing and response assessment in GIST (Ref. Reference Gao, Li, Wang and Ye164). These reflect divergent methodologies rather than disagreement on the trajectory of biomarker-driven care.

Emerging technologies shaping future precision oncology

Artificial intelligence and machine learning are increasingly applied across the precision oncology pipeline, from early drug discovery to clinical biomarker discovery and treatment selection (Ref. Reference Lin, Zhao, Liu, Kang, Zhang, Huang and Yu165). Translational examples include weakly supervised deep-learning methods validated on large-scale whole-slide pathology datasets (Ref. Reference Campanella, Hanna, Geneslaw, Miraflor, Werneck Krauss Silva, Busam, Brogi, Reuter, Klimstra and Fuchs166) and curated precision-oncology knowledge bases such as OncoKB. In October 2021, the FDA recognized a defined subset of OncoKB assertions as valid scientific evidence for supporting the clinical validity of genomic tests (Refs Reference Chakravarty, Gao, Phillips, Kundra, Zhang, Wang, Rudolph, Yaeger, Soumerai, Nissan, Chang, Chandarlapaty, Traina, Paik, Ho, Hantash, Grupe, Baxi, Callahan, Snyder, Chi, Danila, Gounder, Harding, Hellmann, Iyer, Janjigian, Kaley, Levine, Lowery, Omuro, Postow, Rathkopf, Shoushtari, Shukla, Voss, Paraiso, Zehir, Berger, Taylor, Saltz, Riely, Ladanyi, Hyman, Baselga, Sabbatini, Solit and Schultz167, 168). Disease-specific applications further illustrate this translational potential: Guo et al. developed machine-learning models that simultaneously predicted survival and site-specific recurrence in a multi-institutional cervical cancer cohort, demonstrating how such models may complement molecular biomarkers to support individualized risk stratification and surveillance in precision oncology (Ref. Reference Guo, Wang, Wang, Qu, Shi, Meng, Qiu and Hua169). Similarly, deep-learning analysis of routine H&E-stained whole-slide images has shown potential for predicting postoperative metastasis and recurrence in cervical cancer. A hybrid Xception-random forest model achieved an AUC of 0.82, suggesting that digital pathology may provide prognostic information without extensive molecular profiling and help identify patients who could benefit from adjuvant therapy, although prospective clinical validation is still required (Ref. Reference Ye, Zhang, Liang, Lang, Zhang, Zang, Yuan, Tian, Xiao and Yang170). These examples illustrate translational progress without implying that all such systems are independently validated for routine treatment selection.

In immunotherapy, computational integration of multi-omic data is being explored to predict ICI response and characterize the tumour immune microenvironment. The Tumour Immune Dysfunction and Exclusion framework models T-cell dysfunction and exclusion signatures and showed predictive performance across several retrospective cohorts (Ref. Reference Jiang, Gu, Pan, Fu, Sahu, Hu, Li, Traugh, Bu, Li, Liu, Freeman, Brown, Wucherpfennig and Liu171). Despite this promise, clinical translation remains limited by data standardization, model interpretability, generalizability and regulatory uncertainty (Ref. Reference Wang, Xiong, Cui, Duan, Ding, Huang and Wang172).

Single-cell sequencing and spatial transcriptomics are complementary technologies reshaping the understanding of tumour heterogeneity and resistance at unprecedented resolution. Single-cell RNA sequencing resolves clonal architecture, drug-tolerant persister states and immune microenvironment remodelling that bulk sequencing averages out, and is now incorporated as a correlative endpoint in several hundred active clinical trials (Ref. Reference Fuchs, Menzel, Mößinger, Kazdal, Kahles, Budczies and Stenzinger173).

Spatial transcriptomics adds the critical dimension of tissue architecture, mapping gene expression, immune infiltration patterns and genetically distinct clones within their native spatial context. These insights are relevant to understanding immune-evasive niches and optimizing combination immunotherapy strategies (Ref. Reference Liu, Xiao, Wu, Yue, Li, Sun and Yang174).

Integration of single-cell and spatial multi-omic data with computational modelling is expected to identify new predictive biomarkers and inform spatially guided treatment selection, though clinical standardization and cost remain barriers to routine implementation.

These technologies converge on a common goal: enabling truly personalized, biomarker-adaptive combination therapy informed by real-time, multidimensional tumour profiling, in parallel with adaptive and platform trial designs (e.g., NCI-MATCH, TAPUR, ComboMATCH) that accelerate matched-therapy identification outside conventional tumour-type-restricted trials.

Future perspectives

The rapid expansion of tumour-agnostic approvals has further reinforced the centrality of comprehensive molecular profiling in clinical practice. Currently, six pan-cancer biomarkers, NTRK fusions, BRAF V600E, RET fusions, HER2-positive expression, MSI-H/dMMR and TMB-H, have corresponding FDA-approved therapies, with additional emerging candidates including FGFR, ALK, MET, KRAS G12C, NRG1 and homologous recombination deficiency. However, responses to tumour-agnostic therapies are not uniform across histologies; tissue-specific resistance mechanisms, bypass signalling and co-occurring genomic alterations can attenuate efficacy, particularly in pancreatic and microsatellite-stable colorectal cancers (Ref. Reference Myat, Thein, Han, Ahluwalia, Mukherjee and Aung175). This variability underscores that while mutation-driven therapy represents a fundamental advance, tumour lineage and microenvironmental context remain important determinants of response and future biomarker strategies will likely need to integrate multi-omic data, including transcriptomic, proteomic and spatial heterogeneity analyses, to improve predictive accuracy.

The development of living guidelines for tumour-agnostic therapies has been proposed to address the challenge of rapidly evolving evidence and the absence of organ-based frameworks for these cross-histology indications. Such continuously updated guidelines would enable timely incorporation of new approvals, harmonize biomarker testing practices and reduce treatment delays for patients with rare or ultra-rare cancers. In parallel, molecular tumour boards have assumed an increasingly important role in interpreting complex genomic profiles and matching patients to available therapies, including investigational agents through precision oncology trials. These institutional structures, combined with advances in artificial intelligence and bioinformatics, are expected to improve the clinical implementation of precision oncology and bridge the gap between molecular discovery and patient access.

Economically, mAbs and TKIs pose challenges due to high upfront costs, even though their targeted nature can reduce off-target toxicity and improve quality-adjusted survival. Cost-effectiveness concerns are especially relevant when benefits are modest or restricted to narrow patient subgroups. Nevertheless, real-world evidence shows meaningful effectiveness in cancers such as NSCLC, HER2-positive breast cancer and hematologic malignancies, with improved survival, functional outcomes and tolerability.

From an implementation perspective, orally administered TKIs can reduce infusion-related burden, whereas mAbs and ADCs generally require parenteral administration and class-specific monitoring. The main systemic barriers are access to comprehensive molecular profiling, timely pathology reporting, toxicity surveillance and clinician expertise in interpreting complex biomarker data. Addressing these gaps is critical to maximizing the impact of precision oncology.

In the coming years, combinatorial and biomarker-adaptive regimens are expected to dominate, informed by real-world data, molecular profiling and AI-assisted decision tools. Tumour-agnostic approvals may become more common, and guidelines will increasingly reflect molecular classification over histology. Oncologists will need fluency in genomics, bioinformatics and adaptive trial design alongside traditional anatomical and histopathologic expertise.

Ultimately, the future of oncology will rely on the intelligent integration of mAbs and TKIs into multi-modal, adaptive strategies that address tumour evolution and patient-specific context. By combining these agents with next-generation therapies and precision-guided approaches, the field can advance toward durable disease control, functional cures and truly personalized care.

Summary and conclusion

mAbs and TKIs have fundamentally transformed oncology, shifting treatment from empiric cytotoxic therapy toward molecularly guided, precision-based strategies. Their integration into National Comprehensive Cancer Network guidelines reflects both robust clinical evidence and widespread consensus that biomarker-driven therapy is now standard of care across multiple tumour types and disease stages, from preoperative and postoperative settings to advanced metastatic disease.

The impact of targeted agents spans hematologic and solid tumours. In hematologic malignancies, BCR-ABL1 inhibition with imatinib or BCL-2 blockade with venetoclax have substantially improved disease control and survival in molecularly defined settings. In solid tumours, agents such as trastuzumab, osimertinib and alectinib have improved outcomes, while ICIs and tumour-agnostic therapies have extended options for rare or refractory cancers. These successes highlight the importance of molecular selection, central nervous system penetration, resistance management and rational combination strategies.

Despite these advances, challenges persist. Primary and acquired resistance, tumour heterogeneity and microenvironmental influences can limit efficacy. Next-generation inhibitors, combination therapies and earlier intervention in preoperative or minimal-residual-disease settings are essential strategies to overcome these barriers. The rise of tumour-agnostic approvals underscores the critical role of comprehensive genomic profiling and precision diagnostics in everyday oncology practice.

Targeted therapies exemplify the evolution of cancer care, offering durable, life-prolonging and potentially curative options that continue to redefine the standard of treatment across malignancies.