Treatment‐related adverse events of antibody‐drug conjugates in clinical trials: A systematic review and meta‐analysis

Abstract Background The wide use of antibody‐drug conjugates (ADCs) is transforming the cancer‐treatment landscape. Understanding the treatment‐related adverse events (AEs) of ADCs is crucial for their clinical application. We conducted a meta‐analysis to analyze the profile and incidence of AEs related to ADC use in the treatment of solid tumors and hematological malignancies. Methods We searched the PubMed, Embase, and Cochrane Library databases for articles published from January 2001 to October 2022. The overall profile and incidence of all‐grade and grade ≥ 3 treatment‐related AEs were the primary outcomes of the analysis. Results A total of 138 trials involving 15,473 patients were included in this study. The overall incidence of any‐grade treatment‐related AEs was 100.0% (95% confidence interval [CI]: 99.9%–100.0%; I 2 = 89%) and the incidence of grade ≥ 3 treatment‐related AEs was 6.2% (95% CI: 3.0%–12.4%; I² = 99%). Conclusions This study provides a comprehensive overview of AEs related to ADCs used for cancer treatment. ADC use resulted in a high incidence of any‐grade AEs but a low incidence of grade ≥ 3 AEs. The AE profiles and incidence differed according to cancer type, ADC type, and ADC components.


| INTRODUCTION
The emergence of antibody-drug conjugates (ADCs) implied a paradigm shift in the treatment of many solid and hematologic malignancies.To date, 14 ADCs have been approved by the US Food and Drug Administration, the European Medicines Agency, National Medical Products Administration in China, and Japan's Ministry of Health, Labour and Welfare, for the treatment of solid tumors and hematological malignancies (Supporting Information: eTable 1) [1][2][3][4][5][6][7].
In addition, over 100 ADCs are currently being evaluated in clinical trials worldwide [2].ADCs are highly potent biopharmaceutical drugs linking a cytotoxic agent (payload) to a monoclonal antibody via a chemical linker, thus allowing the preferential delivery of toxic payloads to cancer cells while sparing normal cells [1][2][3].
The wide use of ADCs requires a complete understanding of their toxicologic profile to allow the selection of a safe and efficacious dose.Limitations such as target specificity, linker stability, payload delivery, and the payload itself can induce adverse effects, which can be specific or even life-threatening [8][9][10].For instance, several randomized controlled trials (RCTs) demonstrated that ADCs can cause interstitial lung disease, ocular toxicity, serious organ dysfunction, anaphylaxis, severe thrombocytopenia, and neutropenia, as well as gastrointestinal effects [8,11].These adverse events (AEs) remain a significant challenge to the effective clinical application of ADCs.Despite this evidence, there has been no exhaustive overview of the profiles, incidence, and features of ADC-related AEs.There is thus a need to summarize the profiles and incidence of ADC-related AEs using standardized methods, to assist clinicians and researchers to manage these AEs and optimize the trial design of ADCs.
We conducted a systematic review and metaanalysis of published clinical trials reporting treatment-related AEs of ADCs approved by drug administrations worldwide, to provide complete profiles and data on the incidence of ADC-related AEs.We also quantified potential differences in the incidence of AEs across various cancer types, ADC drugs, and ADC components.

| Search strategy and selection criteria
This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines [12].We performed a systematic search of the literature to identify published clinical trials of ADCs that reported treatmentrelated AEs.We searched the terms "antibody-drug conjugates," "cancer," and "clinical trials" in PubMed, Embase, and the Cochrane Library to identify relevant studies published in English between January 2001 and October 2022 (Supporting Information: eTable 2).We also searched the reference lists of relevant review articles manually to identify additional eligible studies.
The inclusion criteria were as follows: (1) prospective clinical trial for cancer treatment between January 2001 and October 2022; (2) treatment with single-agent ADC; (3) ADC drugs approved by any governmental drug administration worldwide; (4) reported overall incidence or tabulated data on treatment-related AEs; and (5) published in English.We excluded conference abstracts that did not contain detailed AE-related data and that had <10 trial participants.The literature search, study selection, and data extraction were performed independently by two authors (G.S. and J.L.).Discrepancies were resolved by discussion with a third reviewer (Z.L.) until consensus was achieved.If multiple articles described the same trial, the article with the most recent and/or comprehensive AE data was used.

| Data analysis
The trial name, first author, year of publication, region, ADC used, ADC component (antibody, linker, and payload), trial phase, cancer type, study design, total number of participants, number of participants in safety analysis, arms and treatment, Common Terminology Criteria for Adverse Events (CTCAE) version, follow-up time, and total number of all-grade AEs were extracted for each study.The definitions of AEs were based on the Medical Dictionary for Regulatory Activities.Any-grade, grade 3 or 4 AEs, and

| Statistical analysis
The effect size in our study was the incidence of AEs, obtained by dividing the number of participants with AEs by the total number of participants.Because the AE incidence did not follow a normal distribution, we performed logit conversion for AE incidence before the meta-analysis [13,14].The pooled incidence with 95% confidence intervals (CIs) was calculated using generalized linear mixed models (GLMMs) [15].When the number of AEs and total number of people included in the study were identical in all arms, the GLMM could not be fitted and a random-effects model with restricted maximum likelihood estimation was used.Subgroup analyses of AE incidence were performed according to ADC drug, linker, payload, target, and tumor type.Heterogeneity between arms was assessed using the χ 2 test and I 2 statistic [16].Heterogeneity was calculated using the Cochrane Q statistic and the I 2 test.Statistical heterogeneity was defined as p < 0.1 and/or I 2 > 50%.Publication bias was evaluated using a modified funnel plot of log odds against the sample size, because a conventional funnel plot against the standard error might be unreliable when the incidence is close to 0% or 100% [17].Egger's test using sample size as a predictor was used to investigate publication bias [18].The metafor and forestplot packages in R v.4.0.4 (www.r-project.org) were used for meta-analysis and forest plot construction, respectively.p < 0.05 was considered statistically significant.1.For all studies, ADC therapy was evaluated in a relapsed, refractory, advanced, or metastatic setting, and most studies reported AEs as secondary outcomes.The cancer types included breast cancer (n = 28), cervical cancer (n = 5), gastric cancer or other solid tumors (n = 10), urothelial carcinoma (n = 12), lung cancer (n = 6), head and neck squamous cell carcinoma (n = 2), lymphoma (n = 34), leukemia (n = 35), and multiple myeloma (n = 6).In total, 138 trials (15,473 patients) and 110 trials (14,183 patients) were included for analyses of the profile and overall incidence of ADCrelated AEs, respectively.

| Overall incidence of AEs
In general, these studies reported >300 types of AEs.A total of 14,320 (92%) among 15,473 participants from 138 studies experienced at least one any-grade AE and 2,695 (20.57%) of 13,101 participants from 98 studies experienced at least one grade ≥ 3 AE.Considering that this study mainly focused on grade ≥ 3 and any-grade AEs, and we restricted ADC-related AEs to the 28 most common, important, or specific AEs.
There was obvious asymmetry in the modified funnel plots for all-grade AEs (Figure 4a) with p < 0.001 in Egger's test; the incidence of all-grade AEs was 97.9% (95% CI: 97.3%-98.3%)after trim and fill correction.No obvious asymmetry was observed in the modified funnel plots for grade ≥ 3 AEs (Figure 4b) (p = 0.665 in Egger's test).Because obvious heterogeneity was observed in the incidence of all-grade and Gade ≥ 3 AEs across arms, we performed subgroup analyses to explore the source of heterogeneity.

| Incidence of special interest AEs related to ADCs
We analyzed AEs of special interest related to ADCs, including ophthalmic toxicity, hepatotoxicity, peripheral neuropathy, severely decreased neutrophil count, left cardiac dysfunction, diarrhea, dermatological toxicity, interstitial lung disease, and pneumonitis.In total, 106 (76.26%) trials reporting the incidence of ADC-related AEs were included.

| Subgroup analysis of overall AE incidence based on ADC components
ADCs are composed of three components: antibody, linker, and payload.The overall incidence of AEs was evaluated according to three types (HER-2, TROP-2, and nectin-4) of antibodies in solid tumors, and the incidence of grade ≥ 3 AEs was evaluated according to nine types of

| DISCUSSION
Several ADCs have been approved by drug administrations and have been widely used in clinical trials and emerging studies worldwide.Despite the increased focus by clinicians and researchers on the safety of ADCs, however, a comprehensive AE profile for ADCs remains to be clearly defined.To the best of our knowledge, the current study was the first to analyze the incidence of ADC-related AEs in studies published to date.The results showed that patients treated with ADCs had a high overall incidence of any-grade AEs but a relatively low incidence of grade ≥ 3 AEs.Several ADC-related special interest AEs were noted, as well as differences in the incidences of any-grade and grade ≥ 3 AEs among different cancer types, ADC types, and ADC components.ADCs are typically composed of an antibody and a cytotoxic payload linked via a chemical linker [8].The specific cytotoxic payload can selectively target cancer cells through delivery of a high-affinity antibody.In principle, ADC-related AEs can be induced by any component of the drug [157,158].The main mechanisms responsible for adverse effects include suboptimal monoclonal antibody specificity, target antigen expression on normal cells, early cleavage of the linker, drug immunogenicity, and binding to Fc and mannose receptors.Off-target effects of cytotoxic payloads associated with unwanted bystander effects are thought to be the primary cause of ADC-related AEs [1,159,160].
To the best of our knowledge, only one previous meta-analysis, including 70 publications, has evaluated the incidence of ADC-related AEs, which concluded that most grade 3 and 4 ADC-associated toxicities were related to the payload [161].However, that study had several limitations.First, it primarily focused on the incidence of grade ≥ 3 AEs related to payload class, rather than summarizing and comprehensively evaluating anygrade AEs.Second, it included agents that were not yet in the market.Third, it did not analyze other ADC components or different populations in depth, and was published 5 years ago.Fourth, it only analyzed some AEs and missed several important ones, such as pneumonitis, hepatotoxicity, and ophthalmic toxicity, which are important with respect to ADCs.There is thus a need for a comprehensive analysis of all common AEs caused by drug administration-approved ADCs, including those previously reported in RCTs [162].
In the present study, the overall incidence of anygrade AEs was 100.0%(95% CI: 99.9%-100.0%)and that of grade ≥ 3 AEs was 6.2% (95% CI: 3.0%-12.4%).Decreased platelet count was the most common anygrade AE (34.9%) and the most common grade ≥ 3 AE (20.1%).Although less likely to be severe at presentation, the incidence of decreased platelet count was relatively high and is thus worth disclosing to patients.Fatigue, peripheral edema, nausea, peripheral sensory neuropathy, decreased neutrophil count, alopecia, and decreased appetite were the next most common any-grade AEs; however, the likelihood of patients experiencing serious manifestations of these AEs is relatively low.Decreased neutrophil count, anemia, leukopenia, and fatigue are common grade ≥ 3 AEs, and patients should not be overly concerned about these AEs related to the cytotoxicity induced by ADC drug linkage.
Our meta-analysis also identified some special interest ADC-related AEs that were less associated with chemotherapy or targeted agents but more common among ADC drugs, including hepatotoxicity, ophthalmic toxicity, dermatological reactions, left cardiac dysfunction, interstitial lung disease and pneumonitis, diarrhea, severely decreased neutrophil count, and peripheral neuropathy.Ophthalmic toxicities caused by ADC drugs are mainly related to the payload (MMAF), which can damage corneal epithelial cells [104,163].Furthermore, ADCs were found to cause hepatotoxicity associated with a calicheamicin payload, increasing the incidence of liver injury or veno-occlusive disease, while MMAFs led to the development of peripheral neuropathy and neutropenic AEs.Moreover, gemtuzumab ozogamicin causes accumulation of antitoxin conjugates in liver cells, resulting in calicheamicin-induced damage [164,165].Trastuzumab deruxtecan comprises an anti-HER2 antibody, tetrapeptide-based cutout linkers, and topoisomerase I inhibitor loads, and interstitial pneumonia caused by this drug is an AE of particular concern compared with other ADC drugs.HER2 expression in the bronchial and fine bronchial epithelium of the lung may be associated with the development of trastuzumab deruxtecan-associated interstitial lung disease [11,166].In a recent study on crab-eating monkeys, immunohistochemical analysis confirmed that trastuzumab deruxtecan was localized mainly in alveolar macrophages but not in lung epithelial cells [167].Nontarget trastuzumab deruxtecan uptake by alveolar macrophages was also observed in animal models, suggesting that lung tissue payload release is involved in the development of interstitial lung disease caused by this drug [167,168].However, further studies are needed to determine the risk factors and mechanisms involved in the development of trastuzumab deruxtecan-associated interstitial lung disease.
This meta-analysis indicated several fatal toxic events of ADCs, with an overall mortality of 0.4%.Interstitial lung disease was the most common cause of ADC-related death, followed by decreased platelet count, dyspnea, and decreased neutrophil count.Recognizing the profile of fatal toxic events is crucial for facilitating their early detection and effectively managing these events.Notably, although clinicians have been familiar with relatively common fatal toxic events such as pneumonitis, some rare fatal events, including diarrhea and dermatological reactions, should also be routinely screened.
Subgroup analyses of the overall AE incidence based on cancer type, ADC type, and ADC components indicated significant differences in all cases.Overall, nine types of cancers were treated with 14 types of ADCs.Among solid tumors, the most frequently reported anygrade AEs were gastrointestinal and hematologic AEs.Among patients with hematological malignancies, the most frequently reported any-grade AEs were ophthalmic and gastrointestinal AEs.grade ≥ 3 AEs accounted for <11% of all AEs for both solid and hematological tumors.The choice of ADC drug target and antibody quality determine the affinity of the ADC for tumor cells, the type of linker determines the stability of the drug, and the choice of payload determines the AEs of ADCs [169].Eleven target types and their monoclonal antibodies were used in these ADCs.The toxicity of ADCs may be related to any of their components.For target analysis, the antibody binds precisely to the antigen in a delivery-specific manner to kill tumor cells, while avoiding binding to normal cells to reduce the resulting toxicity [8].Causes of antibody-induced toxicity include low antibody affinity, insufficient antigen expression on tumor cells, and lack of internalization upon binding.For example, HER-2 protein is predominantly located on the myocardial transverse myocardium, and HER-2 and its downstream signaling pathways are closely associated with cardiac function, further inducing cardiotoxicity.Among the ADCs used, trastuzumab showed a high risk of cardiotoxicity [170,171].Labetuzumab govitecan is an ADC against carcinoembryonic antigen-associated cell adhesion molecule 5 and SN-38, the active metabolites of the antineoplastic drug irinotecan.ADC drugs bind to tumor surface targets through antibodies, and then enter cells through internalization to kill tumor cells.Because Labetuzumab govitecan lacks internalization, this ADC drug will produce certain toxicity after entering Phase I studies [8,172].Most studies have shown that the toxic effects caused by ADCs in clinical settings are mainly related to the payload.
Different ADC components have markedly distinct AE spectra, as observed in the comparison between noncleavable and cleavable linkers.Noncleavable linkers are stable in plasma but have limited efficacy at the target cell, while cleavable linkers are unstable in plasma, leading to off-target toxicity, but have higher efficacy at the target cell [8,173].Our meta-analysis showed that ADCs with cleavable linkers resulted in a higher overall incidence of any-grade and grade ≥ 3 AEs compared with noncleavable linkers.ADCs with both cleavable and noncleavable linkers are associated with a high incidence of gastrointestinal and hematological AEs; however, ADCs with only cleavable linkers mainly cause neurological AEs while ADCs with noncleavable linkers are mainly associated with ophthalmic AEs.We did not carry out any further investigation to determine if specific AEs were more common in particular cancer types (e.g., pneumonitis in lung cancer or colitis in gastrointestinal cancer).
Payload analysis revealed the highest difference in the overall incidence of any-grade and grade ≥ 3 AEs.Most payloads used in ADCs are highly cytotoxic and mediate the AEs of most ADCs.Our results showed that various payload was associated with an increased risk of any-grade AEs, and PBD dimer SG3199 and calicheamicin were associated with increased risks of grade ≥ 3 AEs.SN-38, PBD dimer SG3199, and MMAEs cause adverse respiratory effects, such as pneumonia, skin toxicity, and neurotoxicity, while deruxtecan, camptothecin, SN-38, and calicheamicin mainly cause hematologic toxicity and MMAF causes ophthalmic toxicity.Most payloads can cause digestive and other toxicities, but have low cardiovascular toxicity.
This study had some limitations.First, the diagnosis of AEs was evaluated by the original investigators in the different trials with no standardized or uniform definitions and/or classification.Moreover, the judgment of whether or not an AE was associated with an ADC might be susceptible to bias.There may also be overlap in the extraction process of adverse reaction data, such as elevated liver function indicators, hepatotoxicity, and ADC-related hepatitis.Second, there was heterogeneity among the included trials with regard to targeted drugs, linkers, payload of ADCs, enrolled population, and number of treatment lines.We conducted multiple subgroup analyses to validate the results; however the limited number of included studies meant that we could not analyze several subgroups, for example, whether AEs were caused by a particular ADC component or by a combination of drug and disease factors.Finally, the number of trials of several ADCs was limited.Additional studies with larger sample sizes are therefore needed to confirm our findings.

| CONCLUSIONS
This systematic review and meta-analysis summarized the profile of AEs associated with ADC use and the causes of treatment-related death.ADCs were associated with a high incidence of any-grade AEs but a relatively low incidence of grade ≥ 3 AEs.The type of cancer, type of ADC, and ADC components were associated with varying profiles and incidences of AEs.These findings regarding ADC-related AEs may be useful for clinicians and researchers.

Figure 1
Figure 1 presents a flowchart of the search strategy.The electronic searches yielded 54,800 potentially relevant publications, of which 470 studies were potentially

F I G U R E 4
Funnel plots.(a) All-grade adverse events and (b) grade ≥ 3 adverse events.F I G U R E 5 Incidence of antibody-drug conjugate-related special interest adverse events.(a) Most frequently reported all-grade adverse events and (b) most frequently reported grade ≥ 3 adverse events.CANCER INNOVATION | 363

T A B L E 2
Causes of 129 treatment-related deaths in clinical trials of antibody-drug conjugates.
Main characteristics of the included clinical studies.