Pharmacovigilance study of the association between peripheral neuropathy and antibody–drug conjugates using the FDA adverse event reporting system

Abstract

Antibody–drug conjugates (ADCs) are among the fastest-growing classes of anticancer drugs, making it crucial to evaluate their potential for causing peripheral neuropathy. We analyzed data from the FAERS database (January 1, 2014, to June 30, 2023) using disproportionality and Bayesian methods. We identified 3076 cases of ADC-associated peripheral neuropathy. Our study revealed significant signals for all ADCs (ROR 1.82, 95% CI 1.76–1.89). ADCs with tubulin-binding payloads showed significant peripheral neuropathy signals (ROR 2.31, 95% CI 2.23–2.40), whereas those with DNA-targeting (ROR 0.48, 95% CI 0.39–0.59) and topoisomerase 1 inhibitor (ROR 0.56, 95% CI 0.48–0.66) payloads exhibited non-significant signals. Signals for peripheral sensory neuropathy were 4.83, 2.44, 2.74, and 2.21 (calculated based on IC₀₂₅) for brentuximab vedotin, trastuzumab emtansine, enfortumab vedotin, and polatuzumab vedotin, while signals for peripheral motor neuropathy were 5.31, 0.34, 2.27, and 0.03, respectively. The median time to onset for all ADCs was 127 days (interquartile range 40–457). Tisotumab vedotin had the highest hospitalization rate at 26.67%, followed by brentuximab vedotin at 25.5%. Trastuzumab emtansine had the highest mortality rate ,with 80 deaths (11.96%) among 669 cases. Based on FAERS database, only ADCs with tubulin-binding payloads exhibited significant peripheral neuropathy signals. Brentuximab vedotin and enfortumab vedotin showed similar profiles for peripheral sensory neuropathy and motor neuropathy. Given the delayed time to onset and potentially poor outcomes, ADC-related peripheral neuropathy warrants significant attention.

Introduction

Cancer is a leading cause of mortality globally, responsible for approximately 10 million deaths in 2020, or nearly one in six deaths¹. Cancer treatment includes various approaches such as chemotherapy, immunotherapy, radiation therapy, stem cell therapy, laser therapy, hyperthermia, surgical procedures, and more². Among these, antibody–drug conjugates (ADCs) are one of the fastest-growing classes of oncology drugs in development. Although ADCs aren’t a new concept, their extensive potential to impact the course of clinical practice is well-recognized^2,3. As of December 2021, 14 ADCs were approved to treat both solid and hematologic malignancies, with over 100 ADC candidates remain under clinical trials^4,5. Although ADCs are widely used in cancer treatment, their clinical toxicities, particicularly hematologic, hepatic, and neurological events, considerably hinder their clinical application⁶.

ADCs combine the specificity of monoclonal antibodies with the potency of highly cytotoxic agents, potentially reducing the severity of side effects by preferentially targeting their payload to the tumor site⁷. However, the toxicity and/or side effects resulting from early payload release or ADC binding to normal tissues need to be addressed⁸. Peripheral neuropathy (PN) is a frequent adverse event (AE) that leads to treatment discontinuation and dose reductions⁹. PN is well-known to affect distinct components of the nervous system, from sensory cell bodies in the dorsal root ganglion to the distal axons of primary sensory neurons, causing symptoms such as paresthesia, dysesthesia, and numbness of the hands and feet¹⁰. Importantly, PN can interfere with treatment by compromising adherence and limiting the dosing intensity of long-term therapy, which may reduce efficacy¹¹.

Currently, data regarding the PN profiles of ADC remain limited in real-world clinical practice. The clinical and epidemiological impact of ADC-induced PN can be more accurately assessed using real-world data rather than registration trials¹². Accordingly, this study aimed to investigate the relationships between ADCs and PN and identify the factors influencing these relationships, thereby facilitating future research and clinical surveillance. Based on PN reports in the FDA Adverse Event Reporting System (FAERS) covering the period fromfrom the first quarter of 2014 to the second quarter of 2023, we enumerated PN events, conducted disproportionality analyses to identify PN adverse reactions, and explored potential influencing factors and biological mechanisms related to ADCs.

Results

From January 2014 to June 2023, we extracted 51,336,497 AE reports from the database. After data cleaning, 39,523,594 AE reports were included in the final analysis. Among them, 107,234 reports were submitted for ADC drugs, of which 3076 reports were related to PN (Fig. 1).

Fig. 1.

Process for selecting reports of peripheral neuropathy associated with ADCs from the FAERS database. FAERS Food and drug administration adverse event reporting system, ADC antibody–drug conjugate.

Descriptive analysis

Among 2432 PN cases following ADC administration, the highest number was associated with tubulin binder payloads (n = 2240), followed by topoisomerase 1 inhibitors (n = 121), with DNA-targeting agents accounting for only 71 cases. Table 1 summarizes the clinical features of patients. The group of patients receiving ADCs with tubulin-binding payloads had a higher proportion of males than females (47.54% vs. 34.06%). Stratified by age, patients aged < 65 years accounted for a greater proportion than those aged ≥ 65 years (37.72% vs. 22.01%). Most cases were reported in 2021–2023Q2, reflecting the markedly increased use of ADCs in recent years. Additionally, ADC-associated PN was predominantly reported by healthcare professionals (57.37%), mainly from the United States (24.91%). Hospitalization (21.03%) was the most common ADC-induced outcome, followed by death (11.12%).

Table 1.

Demographic and clinical characteristics of cases associated with peripheral neuropathy receiving different ADCs with specific payloads in FAERS from 2014Q1 to 2023Q2.

Characteristics	Tubulin binders payloads ADCs (N = 2240)	DNA-targeting payloads ADCs (N = 71)	Topoisomerase 1 inhibitors Payloads ADCs (N = 121)	Total ADCs (N = 2432)
Sex
Female	763 (34.06%)	27 (38.03%)	6 (4.96%)	796 (32.73%)
Male	1065 (47.54%)	40 (56.34%)	108 (89.26%)	1213 (49.88%)
Unknown	412 (18.39%)	4 (5.63%)	7 (5.79%)	423 (17.39%)
Age
< 65 years	845 (37.72%)	38 (53.52%)	46 (38.02%)	929 (38.2%)
≥ 65 years	493 (22.01%)	23 (32.39%)	33 (27.27%)	549 (22.57%)
Unknown	902 (40.27%)	10 (14.08%)	42 (34.71%)	954 (39.23%)
Year
2014	87 (3.88%)	2 (2.82%)	0 (0%)	89 (3.66%)
2015	91 (4.06%)	2 (2.82%)	0 (0%)	93 (3.82%)
2016	126 (5.63%)	1 (1.41%)	0 (0%)	127 (5.22%)
2017	91 (4.06%)	3 (4.23%)	0 (0%)	94 (3.87%)
2018	123 (5.49%)	2 (2.82%)	0 (0%)	125 (5.14%)
2019	180 (8.04%)	18 (25.35%)	0 (0%)	198 (8.14%)
2020	312 (13.93%)	7 (9.86%)	5 (4.13%)	324 (13.32%)
2021	395 (17.63%)	8 (11.27%)	11 (9.09%)	414 (17.02%)
2022	410 (18.3%)	16 (22.54%)	39 (32.23%)	465 (19.12%)
2023Q1-Q2	425 (18.97%)	12 (16.9%)	66 (54.55%)	503 (20.68%)
Reported countries
United States	558 (24.91%)	34 (47.89%)	38 (31.4%)	630 (25.9%)
Japan	262 (11.7%)	2 (2.82%)	10 (8.26%)	274 (11.27%)
Canada	163 (7.28%)	0 (0%)	40 (33.06%)	203 (8.35%)
France	159 (7.1%)	1 (1.41%)	1 (0.83%)	161 (6.62%)
Great Britain	149 (6.65%)	16 (22.54%)	19 (15.7%)	184 (7.57%)
Others	712 (31.79%)	17 (23.94%)	10 (8.26%)	739 (30.39%)
Unknown	237 (10.58%)	1 (1.41%)	3 (2.48%)	241 (9.91%)
Type of reporter
Health professional	1285 (57.37%)	50 (70.42%)	32 (26.45%)	1367 (56.21%)
Non-health professional	367( 16.38%)	4 (5.63%)	10 (8.26%)	381 (15.67%)
Unknown	588 (26.25%)	17 (23.94%)	79 (65.29%)	684 (28.13%)
Outcome
Death	249 (11.12%)	9 (12.68%)	15 (12.4%)	273 (11.23%)
Life-threatening	53 (2.37%)	4 (5.63%)	0 (0%)	57 (2.34%)
Hospitalization	471 (21.03%)	30 (42.25%)	34 (28.1%)	535 (22%)
Disability	77 (3.44%)	8 (11.27%)	3 (2.48%)	88 (3.62%)
Other serious	1198 (53.48%)	17 (23.94%)	48 (39.67%)	1263 (51.93%)
Non-serious	192 (8.57%)	3 (4.23%)	21 (17.36%)	216 (8.88%)

FAERS Food and drug administration adverse event reporting system, PN Peripheral neuropathy, ADC antibody–drug conjugate.

Disproportionality and Bayesian analyses

Table 2 presents the associations between ADCs with different payloads and PN. Among all ADC drugs with different payloads, tubulin-binding agents were significantly associated with PN, as evidenced by the highest reported IC and ROR (IC₀₂₅ 1.15; ROR₀₂₅ 2.23). Remarkably, enfortumab vedotin exhibited a stronger association with PN than the other drugs, primarily owing to its high IC and ROR (IC₀₂₅ 1.50; ROR₀₂₅ 2.87). Conversely, ADC drugs with other payloads showed no significant association with PN.

Table 2.

Results of the disproportionality analysis by the payload diversification of marketed ADCs.

Drug class	N	IC	IC025	IC975	ROR	ROR025	ROR975
Total	3080	0.87	0.81	0.91	1.82	1.76	1.89
Tubulin binders Payloads ADCs	2836	1.21	1.15	1.26	2.31	2.23	2.40
Brentuximab vedotin	1251	1.53	1.44	1.60	2.90	2.74	3.06
Trastuzumab emtansine	918	1.09	0.98	1.17	2.13	1.99	2.27
Polatuzumab vedotin	299	0.27	0.08	0.41	1.21	1.08	1.35
Enfortumab vedotin	331	1.68	1.50	1.81	3.20	2.87	3.58
Tisotumab vedotin	37	1.68	1.13	2.07	3.20	2.29	4.45
DNA-targeting payload ADCs	87	− 1.07	− 1.42	− 0.81	0.48	0.39	0.59
Gemtuzumab ozogamicin	19	− 2.05	− 2.82	− 1.51	0.24	0.15	0.38
Inotuzumab ozogamicin	66	− 0.56	− 0.97	− 0.27	0.68	0.53	0.86
Loncastuximab tesirine	2	− 1.02	− 3.61	0.37	0.49	0.12	1.98
Topoisomerase 1 inhibitors payloads ADCs	157	− 0.83	− 1.10	− 0.64	0.56	0.48	0.66
Trastuzumab deruxtecan	59	− 1.21	− 1.64	− 0.90	0.43	0.33	0.56
Sacituzumab govitecan	98	− 0.54	− 0.88	− 0.30	0.69	0.56	0.84

ADCs antibody–drug conjugates, N number of records, IC information components, ROR reporting odds ratio, IC₀₂₅/IC₉₇₅ lower/upper limit of 95% confidence interval for IC, ROR₀₂₅/ROR₉₇₅ lower/upper limit of 95% confidence interval for ROR.

Comparison of toxicity profiles of ADCs with tubulin-binding payloads

Brentuximab vedotin showed the broadest spectrum of PN AEs, with 16 PTs detected as signals ranging from hypotonia (IC₀₂₅ = 0.20) to peripheral motor neuropathy (IC₀₂₅ = 5.31) (Fig. 2). Conversely, 14 PTs, ranging from muscular weakness (IC₀₂₅ = 0.24) to PN (IC₀₂₅ = 2.97), were significantly associated with trastuzumab emtansine treatment. Nine PTs were significantly associated with polatuzumab vedotin, encompassing a spectrum of AEs ranging from peripheral motor neuropathy (IC₀₂₅ = 0.03) to toxic neuropathy (IC₀₂₅ = 3.06). Four PTs were significantly associated with enfortumab vedotin, and two PTs were significantly associated with tisotumab vedotin. Among these, PN emerged as the strongest signal, with IC₀₂₅ values of 4.13 and 2.88 for enfortumab vedotin and tisotumab vedotin, respectively.

Fig. 2.

Peripheral neuropathy signal profiles of different ADCs. ADCs antibody–drug conjugates, BV brentuximab vedotin, TE trastuzumab emtansine, EV enfortumab vedotin, PV polatuzumab vedotin, TV tisotumab vedotin.

Time to onset (TTO) of PN for ADCs with tubulin-binding payloads

The median TTO of ADC-related PN was 127 (interquartile range [IQR] 40–457) days. The median TTO of PN was 56 (IQR 20–195) days for brentuximab vedotin, 53 (IQR 13–468) days for trastuzumab emtansine, 62.5 (IQR 16.5–290.3) days for enfortumab vedotin, 23 (IQR 7–56) days for polatuzumab vedotin, and 83 (IQR 13–259.5) days for tisotumab vedotin (Fig. 3). Almost one-fifth (20.32%) of PN cases occurred within the first month, with nearly half (49.44%) occurring within the initial six months.

Fig. 3.

Time to onset of ADC-related peripheral neuropathy. ADC antibody–drug conjugate, BV brentuximab vedotin, TE trastuzumab emtansine, EV enfortumab vedotin, PV polatuzumab vedotin, TV tisotumab vedotin.

Fatality and hospitalization rates of ADCs with tubulin-binding payloads

To evaluate the prognosis following the administration of ADCs with tubulin-binding payloads, we analyzed the fatality and hospitalization rates associated with PN, considering various tubulin-binding payloads as shown in Fig. 4. Based on collected data, the chi-square and Fisher’s exact tests revealed significant differences in fatality rates (P < 0.001). Although the examined ADCs were associated with low mortality rates, trastuzumab emtansine exhibited the highest risk of death, accounting for 80 deaths (11.96%) among 669 cases. Tisotumab vedotin-induced PN had the highest hospitalization rate (26.67%), probably due to the limited sample size, followed by brentuximab vedotin-related PN (25.5%).

Fig. 4.

The proportion of deaths, life-threatening events, and hospitalizations for ADC-associated peripheral neuropathy. ADC antibody–drug conjugate, BV brentuximab vedotin, TE trastuzumab emtansine, EV enfortumab vedotin, PV polatuzumab vedotin, TV tisotumab vedotin.

Discussion

To the best of our knowledge, the current study represents the first and most extensive compilation describing disparities in vulnerable populations, onset times, and adverse outcomes associated with ADC-related PN in real-world clinical practice, based on the FAERS pharmacovigilance database. Herein, ADCs with tubulin-binding payloads were associated with PN and exhibited diverse characteristics.

Most ADC-induced AEs can be attributed to the payload¹³. ADCs with the same class of linkers/payloads typically exhibit similar toxicity profiles and maximum tolerated doses, irrespective of the target antigen and the level of antigen expression in healthy tissues¹⁴. The prevalence of ADC-induced grade 3/4 toxicities is consistent with their payload class, as demonstrated by a review discussing clinical ADC data¹⁵. After ADC dosing, the released payload quickly enters systemic circulation, where premature deconjugation of the payload results in plasma exposure to the free payload (e.g., due to inadequate linker stability)^16. It is estimated that only 0.1% of the injected ADC dose reaches the intended diseased cell population, whereas the majority of the administered dose is metabolized "off-site" within non-targeted healthy cells, resulting in unintended toxicity^17,18. Lipophilic payloads have high permeability across plasma membranes, resulting in payload release into non-targeted cells, which could lead to unwanted cytotoxicity. Over 80% of approved ADCs possess lipophilic and cleavable linkers¹⁹. In addition to the released payload entering non-targeted cells through passive diffusion across plasma membranes, non-specific endocytosis of intact ADC may contribute to the off-site payload delivery²⁰.

Because small-molecule payloads typically employ the same mechanism of action as traditional anticancer chemotherapy agents following release from a monoclonal antibody, the payload can induce the same typical chemotherapy-related toxicities, including hematologic and non-hematologic AEs¹⁵. PN is a notable dose-limiting, off-target toxicity associated with ADCs with tubulin inhibitor payloads in conjunction with cleavable linkers such as monomethyl auristatin E (MMAE), DM1(N2′-Deacetyl-N2′-(3-mercapto-1-oxopropyl) maytansin); DM4(N2′-Deacetyl-N2′-(4-mercapto-4-methyl-1-oxopentyl) maytansine)^21–24. Tubulin-inhibiting chemotherapeutic agents, including taxanes and vinca alkaloids, have been associated with PN, a common AE²⁵. It is speculated that PN can be attributed to peripheral axonopathy caused by the free payload released into the systemic circulation²⁰. Consistent with previous reports, our findings revealed that ADCs with tubulin-binding payloads showed significant PN signals when compared with those of DNA-targeting and topoisomerase 1 inhibitor payloads.

Auristatins (including MMAE and monomethyl auristatin F [MMAF]) and myotanninoids (including DM1 and DM4) inhibit microtubule assembly, leading to cell cycle arrest²⁶. For all ADCs, the rate of G3/4 toxicity in PN is reportedly low; however, this is most commonly observed with ADCs with an MMAE payload (6.5%)¹⁵. Vedotin is a term used to describe MMAE and its linkage to the antibodies. MMAE is a synthetic drug utilizing the auristatin structure and derived from dolastatin, a natural product²⁷. MMAE can induce PN, a well-known side effect of tubulin-binding agents²⁸. Mechanistically, MMAE binds to tubulin extensively, inhibits microtubule (MT), and induces severe inhibition of MT dysregulation by blocking tubulin polymerization²⁹. Moreover, the tubulin-bound MMAE suppresses MT polymerization, leading to G2–M phase growth arrest and cell apoptosis^30,31. MMAE-mediated inhibition of MT-dependent axonal transport can induce susceptibility to PN, mainly because of the lengthy axonal projections and the crucial role played by the MT network in sustaining long-distance axonal transport between neuronal cell bodies and remote nerve endings³². The release of free drugs owing to the early cleavage of the linker may result in more widespread toxicity. MMAE conjugates linked to a protease-cleavable linker are less stable than those linked to other linkers, leading to the systemic release of free drugs³³. Regardless of the target antigen, PN has been consistently associated with all conventional MMAE ADCs^9,14. According to a phase II study assessing polatuzumab vedotin, the incidence of grade ≥ 2 PN in patients with indolent non-Hodgkin lymphoma was 55–72%³⁴. In a phase III study, 112 patients (67%) in the brentuximab vedotin group were found to develop treatment-emergent PN³⁵. PN was the most common treatment-related AE (TRAE; 38%) documented in a phase I clinical trial of enfortumab vedotin and occurred in 50% and 46.3% of patients in phase II and III clinical trials, respectively^36–38.

In preliminary clinical trials, the low incidence of AEs (such as sensory neuropathy, diarrhea, and vomiting) typically observed with maytansinoids is consistent with minimal systemic exposure to DM1³⁹. PN was not detected among the most common adverse drug events associated with trastuzumab emtansine⁴⁰. Similarly, the signal intensity of trastuzumab emtansine was relatively low in the current study. This finding could be attributed to the unique and stable thioether linker used to conjugate DM1 to trastuzumab. The extent of ADC-induced toxicity may be influenced by components such as the antibody, linker, and payload of the ADC²⁰. Apart from linker-drug instability leading to the premature release of cytotoxic drugs (payload) into the bloodstream, the uptake/trafficking of intact ADCs can also occur through both receptor-dependent and receptor-independent mechanisms (non-specific endocytosis). This dual process may contribute to the off-target toxicity in healthy cells¹⁹, clarifying the difference in signal values between the same payload and linker observed in the current study, with enfortumab vedotin exhibiting the highest (ROR₀₂₅ = 2.87) and polatuzumab vedotin exhibiting the lowest (ROR₀₂₅ = 1.08) values.

Herein, brentuximab vedotin and enfortumab vedotin had similar signals in peripheral sensory neuropathy (PSN) and peripheral motor neuropathy (PMN), whereas trastuzumab emtansine and polatuzumab vedotin had significantly higher signal values in PSN than in PMN. These results do not completely align with those of previous clinical drug trials. In a pivotal phase II trial of brentuximab vedotin, incidence rates of PSN and PMN of any grade were 42% and 11%, respectively. Treatment discontinuation was undertaken in approximately 20% of patients, with PSN (6%) and PMN (3%) deemed the most common AEs. PSN (13%) was one of the most common causes of dose delays (8%)⁴¹. In the phase III AETHERA study, the incidence of PSN and PMN of any grade was 56% and 23%, respectively, in patients with Hodgkin’s lymphoma who received brentuximab vedotin as consolidation therapy, whereas the incidence of grade ≥ 3 PSN and PMN was 10% and 6%, respectively³⁵. Additionally, PSN was found to be more prevalent (44%) than PMN (14%) in phase II clinical trials of enfortumab vedotin³⁷. In a phase III clinical trial, PSN of any grade occurred in 43.9% of patients, while PMN occurred in 7.4%; PSN was the most common TRAE leading to dose reduction (7.1%), interruption (15.5%), and withdrawal (2.4%)³⁸. Preliminary clinical data indicate that the incidence of PSN following treatment with brentuximab vedotin and enfortumab vedotin is substantially higher than that of PMN. In this phase II randomized study, patients received rituximab plus polatuzumab vedotin every 21 days, and the incidence of PSN and PMN of any grade was 23% and 10%, with grade 3–4 accounting for 8% and 3%, respectively⁴². In the current study, the PSN signal for polatuzumab vedotin was higher than that for PMN.

The median TTO of PN for brentuximab vedotin was 8 weeks, which is shorter than that reported in previous clinical trials. Considering the brentuximab vedotin group, the median TTO PN events in phase I, II, and III clinical trials were 9, 12.4, and 13.7 weeks, respectively^30,35,41. Similarly, for enfortumab vedotin, the median TTO of PN was 2.43 (0.03–7.39) months and 2.694 (0.03–11.99) months in phase II and III clinical trials, respectively^37,38. In the current study, the median TTO was approximately 2.08 weeks. The median TTO of PN was 3.1 months (IQR 1.8–4.4) and 8.7 weeks (IQR 3.0–14.1) in phase II trials of metastatic cervical cancer or metastatic solid tumors, respectively^43,44. The TTO detected in the current study was 83 days (13–259.5), ranging between the previously determined values. Herein, brentuximab vedotin was associated with relatively high rates of hospitalization (25.5%) and death (11.2%). PN either resolved or showed some improvement (of one grade or more) in 80% of patients in phase II trials, while 50% of patients experienced complete resolution of all PN events⁴¹. In phase III trials, PN led to the discontinuation of brentuximab vedotin treatment in 38 (23%) patients and required dose modification (dose reduction or delay) in 51 (31%) patients³⁵. Trastuzumab emtansine was associated with the highest mortality (11.96%), which differs substantially from that observed in a phase I clinical trial report, in which no PN or deaths occurred⁴⁵. Owing to the delayed TTO and potentially poor outcomes, monitoring and following up with patients more frequently and for a prolonged period may be necessary.

Our study had several limitations. First, we employed a SRS for qualitative research, which does not allow for direct safety comparisons, quantification of associations, or calculation of incidence rates because of missing denominator data³⁴. Second, data derived from the SRS are less dependable than those gathered from clinical trials and cohort studies. Identifying and reporting AEs within SRS are subject to less stringent control. Third, due to the lack of baseline characteristics such as prior chemotherapy, comorbidities, and baseline neuropathy, the FAERS data may suffer from confounding issues for PN associated with ADCs. Fourth, brentuximab vedotin accounted for a high proportion of the reported data (44.11%), with only 1.30% of tisotumab vedotin-related cases reported. The potential for considerable bias due to limited reporting should be considered. Finally, data mining revealed imperfect reporting with inaccuracies and incomplete entries, potentially causing analytical bias. Despite the limitations of the FAERS database, our findings highlight essential ADCs-associated PN aspects, guiding future rigorous research for result validation.

Conclusion

Based on the FAERS database, we profiled PN for various ADC drugs, elaborating on the occurrence, clinical features, and prognosis. Only ADCs with tubulin-binding payloads showed significant PN signals. Brentuximab vedotin and enfortumab vedotin have similar effects on peripheral sensory and motor neuropathies. Owing to the prolonged TTO and potentially poor outcomes, ADC-related PN should be carefully monitored.

Materials and methods

Study design and participants

In order to evaluate the PN AEs associated with ADCs in real-world settings, this real-world observational retrospective study performed a disproportionality analysis based on the FAERS database using data from the first quarter of 2014 to the second quarter of 2023. The FAERS database is a typical spontaneous reporting system (SRS) that collects data from sources, such as AE reports, medication error reports, and product quality complaints, collating AEs reported by healthcare professionals, consumers, and manufacturers. The reports include information on patient demographics, medication use, AEs, indications, outcomes, and report sources. All AEs are coded according to the Medical Dictionary for Regulatory Activity (MedDRA). The FAERS database allows for signal detection and quantification of the association between a drug and its reactions. Given that data in FAERS are anonymized and publicly available, the requirement for obtaining informed consent and institutional review board approval was waived.

Data processing

Essential variables, such as PRIMARYID, CASEID, SEX, DRUGNAME, ROLE_COD, and preferred terms (PT), were extracted from the different data files in the database. The FAERS database invariably includes duplicate reports submitted by various individuals and institutions. Therefore, duplicates were removed to reduce false-positive and false-negative signals by employing a simple but widespread method, i.e., variable matching⁴⁶. The variable matching method involves matching key variables in two reports. If the key variables were the same, the two reports were considered duplicates. Although key variables can be customized, they generally include report IDs, patient details (such as sex, birth date), and suspected drugs. As recommended by the FDA, only the most recent report should be used. Therefore, we selected PRIMARYID, CASEID, CASEVERSION, and FDA_DT as key matching variables. The procedure involved selecting the latest FDA_DT when the PRIMARYIDs were identical and choosing the largest CASEID and CASEVERSION when the FDA_DT and PRIMARYID were identical. Drugs were categorized into four groups: primary suspect (PS), secondary suspect (SS), concomitant (C), and interacting (I). Concomitant-associated records were excluded to obtain better signal intensity, as adopted by the World Health Organization (WHO) Uppsala Monitoring Centre. Since FAERS has two medication-related variables, DRUGNAME and PROD_AI, both generic and brand names were used to identify ADCs in the database.

Toxicities associated with ADC administration typically align with side effects related to the specific payload type. Therefore, we conducted a stratified analysis based on the payload categories. The search was performed using the words: gemtuzumab ozogamicin/Mylotarg, brentuximab vedotin/Adcetris, ado-trastuzumab emtansine/Kadcyla, inotuzumab ozogamicin/Besponsa, polatuzumab vedotin/Polivy, enfortumab vedotin/Padcev, fam-trastuzumab deruxtecan/Enhertu, sacituzumab govitecan/Trodelvy, tisotumab vedotin/Tivdak, and loncastuximab tesirine/Zynlonta. All PN AEs were coded as PTs according to MedDRA version 25.1.

Statistical analysis

Disproportionality analysis was performed using two data mining methods, proportional reports reporting odds ratios (ROR) and Bayesian confidence propagation neural networks of information components (IC)^47,48, with all other drugs/events recorded in FAERS as a comparator. For the ROR, the lower limit of the 95% confidence interval (CI) of ROR (ROR₀₂₅) > 1 with at least three cases indicated a significant signal. For IC, the lower end of the 95% CI of IC (IC₀₂₅) > 0 suggested a significant signal. Statistical shrinkage transformation was applied to obtain robust results⁴⁷. The statistical formula for shrinkage transformation is as follows:

$$\text{ROR}=\frac{{\text{N}}_{\text{observed}}+0.5}{{\text{N}}_{\text{excepted}}+0.5}$$

$$\text{IC}={\text{log}}_2\frac{{\text{N}}_{\text{observed}}+0.5}{{\text{N}}_{\text{expected}}+0.5}$$

$${\text{N}}_{\text{expected}}=\frac{{\text{n}}_{\text{drug}}\ast {\text{n}}_{\text{event}}}{{\text{n}}_{\text{total}}}$$

N_expected: number of records expected for the selected drug-AE combination. N_observed: observed number of records for the selected drug-AE combination. N_drug: the total number of records for the selected drug. N_event: the total number of total records for the selected AE. N_total: the total number of records in the database.

Data analyses were performed using R Software (https://www.r-project.org/; version 3.6.1), and statistical significance was defined as p < 0.05.

Author contributions

Y.C. and X.R. drafted the manuscript. Y.W.: Research planning; data curation; resources; review and editing; visualization. All authors participated in the data analysis and interpretation, manuscript revision, and final approval of the submission.

Funding

This work was funded by the CAMS Innovation Fund for Medical Sciences (CIFMS) (supported by the Special Research Fund for Central Universities, Peking Union Medical College, 2022-I2M-C&T-B-069) and the program of Beijing Hope Run Special Fund of Cancer Foundation of China (LC2020A17).

Data availability

The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request. The raw data can be obtained from the FAERS database at the following link: FAERS Quarterly Data Extract Files (fda.gov).

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Ethical approval was not sought because the study involved only analysis of data obtained from a public database. Informed consent to participate: not applicable.