Antidrug Antibodies Against Immune Checkpoint Blockers: Impairment of Drug Efficacy or Indication of Immune Activation?

Any biologic agent administered to humans is likely to be recognized by the immune system that induces a humoral response resulting in the generation of antidrug antibodies (ADAs; Fig. 1; ref. 1). These antibodies may bind to the drug and cause loss of activity via different mechanisms, such as blockade of the drug target binding (neutralizing antibodies) or by increasing the clearance of the ADA-drug complex, resulting in reduced drug concentrations in the plasma (2). Furthermore, ADAs may induce toxicities by the generation of an immune response to the ADA–drug complex, with infusion-related reactions being the most frequent events (3).

A range of factors can influence the development of this immunogenicity, such as the origin and structure of the drug, impurities with adjuvant activity, route, dose, and frequency of administration, immunomodulatory properties of the therapeutic product, formulation of recombinant therapeutic proteins, aggregates formed by shaking during preparation shipment, the patient's immunologic and genetic status, and concomitant treatment (4). By their nature, ADAs are a heterogeneous group of antibodies with different isotypes (IgM, IgG, IgE, or IgA). ADA development seems to be an early event, occurring during the drug infusion, and in some cases as early as the first dose (5). Initially there is a low-affinity and nonneutralizing IgM response, followed by higher affinity and neutralizing antibodies mostly IgG1 and IgG4. In patients treated for autoimmune or inflammatory diseases, the titers of ADAs, as well as the persistence of ADAs throughout the treatment, were associated with clinical relevance (6). Adding to the difficulty of interpretation is the fact that there are many testing methods, each with different levels of sensitivity and cut-off points (6). In this context of the difficulty of clinical interpretation of ADA induction, caution must be exercised when comparing ADA incidence between studies and the potential relevance for clinical outcome.

In many tumor types, immune checkpoint inhibitors are now part of the standard of care. Humanized and fully human antibodies against PD-L1 or PD-1 have had the most success to date, with five drugs approved (nivolumab, pembrolizumab, atezolizumab, durvalumab, and avelumab). Over a very short period, many trials with very similar designs have been conducted with these drugs and, although their targets are similar, there are some discrepancies in the results that are difficult to explain. As an example, in the subgroup of non–small cell lung cancers (NSCLCs) expressing PD-L1 on at least 50% of the tumors cells, pembrolizumab significantly improved both progression-free survival and overall survival (OS) over platinum-based chemotherapy, whereas nivolumab and durvalumab failed to demonstrate superiority (7–9). Similarly, in second-line therapy in platinum-refractory advanced urothelial carcinoma, pembrolizumab was associated with significantly longer OS than chemotherapy, but atezolizumab failed to reach this endpoint in a similar scenario (10, 11). Among the hypotheses put forward to explain these discrepancies, those highlighted include, patient selection, tumor phenotype, dose, and schedule selection, comedication such as steroids, or host characteristics (microbiota in particular; ref. 12). It is also feasible that drug immunogenicity also contributes to these conflicting results given that one of the main concerns in ADA development is their clinical relevance and impact on drug efficacy.

ADAs have been analyzed in various medical fields, with oncology being a particular point of focus (13). Unfortunately, fewer than 50% of published oncology trials have analyzed the possible effects of ADAs on pharmacokinetics, efficacy, or safety parameters (14). The incidence of ADAs induced by immune checkpoint inhibitors varies largely between drugs, as summarized in Fig. 2 (15–26). Antinivolumab antibodies were present in 11.2% of the 2,085 patients analyzed using the electrochemiluminescence (ECL) immunoassay (17). Interestingly, in patients who were treated with nivolumab and ipilimumab, the presence of antinivolumab antibodies was between 23.8% to 37.8%, whereas anti-ipilimumab antibodies ranged from 4.1% to 8.4%. Agrawal and colleagues (27) demonstrated the development of antinivolumab antibodies in 12.7% of 1,086 patients with advanced melanoma or lung cancer from six phase II/III clinical studies using the ECL immunoassay. For atezolizumab, pembrolizumab, durvalumab, and avelumab, the maximum ADA-positive rates reported were 54.1%, 2.1%, 2.9%, and 5.9%, respectively (15, 22, 23, 25).

It is well known that the intrinsic immunogenicity of mAbs has been progressively reduced from murine, chimeric to humanized and fully human antibodies, but even these latter two types of antibodies cannot prevent the ADA development (28). Theoretically, fully human antibodies should have significantly lower immunogenicity compared with humanized, but today this principle is still a matter of debate because it could not be widely demonstrated in nononcologic settings (29). This postulate is also not valid for the immune checkpoint inhibitors considering the frequency of ADAs reported. Humanized Fc-engineered atezolizumab presented higher ADAs frequency (54.1%) compared with the fully human ipilimumab (5.4%), avelumab (5.9%) and durvalumab (2.9%). However, the humanized pembrolizumab reported an ADA incidence of 2.1% compared with 11.2% with the fully human nivolumab. These results demonstrate, with the limitations that all these studies used tests with different sensitivity, that the use of humanized and/or fully human mAbs in patients with cancer is not systematically associated with reduced induction of ADAs, as described previously in patients with inflammatory diseases (6).

In a study of single-agent ipilimumab (anti–CTLA-4 antibody) in 31 patients treated for melanoma, the 26% of patients who developed ADAs by a bead-based assay, reflecting in this case anti-ipilimumab antibodies, had significantly shorter OS, compared with the ADA-negative patients (5). The authors defined ADA positivity based on both absolute levels and increase from baseline. Ipilimumab effectiveness is dependent on the dose and the corresponding serum concentration; however, this study did not demonstrate an association between ADA positivity and low serum levels of ipilimumab. Only free forms of ADAs, as well as ipilimumab, were detected because dissociation of ADA–drug complexes was not performed. Contrary to ipilimumab, PD-1 and PD-L1 inhibitors do not have a clear certified dose–response association (30).

While in patients treated with nivolumab, pembrolizumab, durvalumab, or avelumab, there was no evidence that ADAs had clinical consequences (21, 23, 25, 27), atezolizumab is the only checkpoint inhibitor with an expanded analysis of ADAs. The phase III OAK trial compared second-line atezolizumab to docetaxel in 850 patients with NSCLC (31). In a subset analysis of 565 patients, 30% tested positive for ADAs at any postdose time point, and 21% after 4 weeks. These patients had a 25% higher drug clearance compared with ADA-negative patients. The subset of ADA-positive patients did not have increased toxicity; however, an exploratory analysis suggested an impact on OS, with OS in ADA-positive patients similar to that in docetaxel-treated patients [ADA-positive subgroup HR = 0.89 (95% confidence interval (CI), 0.61–1.30); ADA-negative subgroup HR = 0.68 (95% CI, 0.55–0.83); Fig. 3; ref. 16]. These data are not in line with the sister phase II study POPLAR that also compared second-line atezolizumab to docetaxel in 287 patients with NSCLC (32). When comparing the response rate in a specific subgroup analysis of atezolizumab arm according to ADA status, it was found an ORR of 20.5% (95% CI, 12.0–31.6) in the ADA-positive subgroup (n = 73), compared with 9.7% (95% CI, 3.6–19.9) in the ADA-negative population (n = 62). This better response in the ADA-positive subset seems to translate in a progression-free survival benefit, which was 4.1 months (95% CI, 2.7–5.7) in ADA-positive versus 2.7 months (1.5–4.2) in ADA-negative patients (15). A higher ORR in ADA-positive patients was also found in the phase II FIR trial which evaluated the efficacy and safety of atezolizumab in a PD-L1–selected population with NSCLC. The analysis of 31 patients in cohort 1 revealed an ORR of 31.3% in ADA-positive patients compared with 20% in the ADA-negatives (15, 33). In the phase II single-arm IMvigor210 trial of atezolizumab in patients with locally advanced and metastatic urothelial carcinoma who had progressed (cohort 2) or who were cisplatin-ineligible (cohort 1), 42% of the 275 patients, and 48% of 111 patients respectively, were positive for ADAs (16). Even though these patients also had lower systemic exposure of atezolizumab, it had no impact on efficacy, because a subgroup analysis of cohort 2 found an ORR of 19.3% (95% CI, 12.5–27.8) in 114 ADA-positive patients compared to an ORR of 15.5% (95% CI, 10.3–22.1) in 161 ADA-negative patients (15).

While these conflicting results call for prospective evaluations, it could be hypothesized that the difference in clinical efficacy between ADA-positive and negative patients observed in the mentioned trials may explain, at least partly, the lower atezolizumab response rate in advanced lung and urothelial carcinoma trials compared with the other immune checkpoint inhibitors. In the phase III OAK trial, the objective response rate (ORR) for atezolizumab was 14%, which is slightly lower than in other studies in NSCLC, including nivolumab in the phase III CheckMate 017 (20%) and CheckMate 057 (19%), or with pembrolizumab in the phase II/III Keynote-010 (18%; refs. 31, 34–36). Similarly, in platinum-refractory advanced urothelial carcinoma trials, the ORR with atezolizumab was 13.4% (phase III IMvigor211 trial), which was lower compared with pembrolizumab (21.1% in the phase III KEYNOTE-045), nivolumab (19.6% in the phase II CheckMate 275), durvalumab (17.8% in the phase I/II NCT01693562), or avelumab (18.2% in the phase Ib JAVELIN; refs. 10, 11, 37–39). With respect to cisplatin-ineligible patients, a similar tendency was observed for atezolizumab which had a lower ORR (23% in IMvigor210 cohort 1) compared with pembrolizumab (27% in the phase 2 KEYNOTE-052; refs. 40, 41). It should be highlighted that each patient can develop ADAs with different effects (blocking the drug or not). In addition, one could see the development of ADAs as a surrogate marker of the immune system activation. Even though these provocative comparisons are not supported statistically, all the above clinical observations can provide the background for possible hypotheses of future prospective studies evaluating the clinical implication of ADA status when using immune checkpoint inhibitors. Otherwise, one of the most worrying clinical relevance of ADAs is the relationship with toxicity, mostly with infusion-related reactions. High titers of IgE ADAs, after the first drug exposure, can develop a hypersensitivity reaction mediated by degranulation of histamine (42). However, an increasing frequency of infusion-related reactions was not reported for the ADA-positive patients treated with immune checkpoint inhibitors, revealing that this event may also be induced by a nonantibody-mediated mechanism (non–ADA-dependent), such as the cytokine release syndrome (3).

It could be hypothesized that the combination of immune checkpoint inhibitors with chemotherapy may theoretically decrease the generation of ADAs due to the immunosuppressive effect of chemotherapy. This was demonstrated in patients with rheumatoid arthritis concomitantly treated with methotrexate and adalimumab who had a reduced rate of ADA development (43). It is important to consider that low-dose injection of methotrexate may have a different immunomodulatory role compared with doses used for cancer treatment. The phase III Impassion130 trial, which combined atezolizumab and nab-paclitaxel in advanced triple-negative breast cancer, supports this hypothesis. It was demonstrated a low incidence of ADA development (13%) among 434 tested patients (15). Discordant results come from the phase 3 IMpower150 study that evaluated chemotherapy (carboplatin and paclitaxel) plus bevacizumab with or without atezoliazumab in 800 patients with nonsquamous NSCLC (arms B and C; ref. 44). ADA positivity against atezolizumab, among 365 ADA-evaluable patients, was 36% after the first or subsequent doses. An exploratory analysis on efficacy according to ADA status was performed. The hazard ratio for OS in the ADA-positive subgroup was 0.69 (95% CI, 0.44–1.07) comparing with 0.64 (95% CI, 0.46–0.90) in the ADA-negative subgroup (Fig. 4; ref. 16). The subset of ADA-positive patients appeared to have similar efficacy as compared with the ADA-negative group, but the confidence intervals in ADA-positive subgroup were wide. Some strategies to reduce ADA formation by immune tolerance have been proposed outside cancer treatment, such as increasing the dose and frequency of the therapy or adding corticosteroids (45, 46). This last strategy would not seem to be a possible alternative in cancer treatment since the immunosuppressive activity of corticosteroids may reduce the antitumor activity and clinical efficacy of immune checkpoint inhibitors (47, 48). The combination of chemotherapy plus immunotherapy has been established as a standard of care in untreated patients with advanced NSCLC. Nonetheless, studies supporting this do not report data on ADA incidence, and their clinical impact is unknown in this setting, highlighting a gap in our knowledge.

Despite the widespread use of immune checkpoint inhibitors in oncology, there is a lack of information concerning their immunogenicity and its implications. Immunogenicity is very heterogeneous across immune checkpoint inhibitors, and this can be explained by the absence of standardized methodology. Although expanded reports of ADAs were made for atezolizumab, the data are conflicting and additional analyses from published as well as prospective trials are required. A better understanding of the clinical impact of ADAs may help to define therapeutic strategies, for example, by modifying the drug dose in the event of ADA development, or adding chemotherapy to prevent ADA formation as was demonstrated in rheumatology. For now, clinical implications of ADAs against immune checkpoint inhibitors remain to be elucidated.

B. Besse reports receiving commercial research grants from Abbvie, Amgen, AstraZeneca, Biogen, Blueprint Medicines, BMS, Celgene, Eli Lilly, GSK, Ignyta, IPSEN, Merck KGaA, MSD, Nektar, Onxeo, Pfizer, Pharma Mar, Sanofi, Spectrum Pharmaceuticals, Takeda, Tiziana Pharma. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Enrico, B. Besse

Development of methodology: D. Enrico

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Enrico

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Enrico, E. Karamouza, B. Besse

Writing, review, and/or revision of the manuscript: A. Paci, N. Chaput, B. Besse

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Enrico

The authors thank Sarah MacKenzie for English language edition. The authors received no specific funding for this work. D. Enrico was the recipient of a DUERTECC/EURONCO grant for 2018–2019. Sponsored Research at Gustave Roussy Cancer Center: Abbvie, Amgen, AstraZeneca, Biogen, Blueprint Medicines, BMS, Celgene, Eli Lilly, GSK, Ignyta, IPSEN, Merck KGaA, MSD, Nektar, Onxeo, Pfizer, Pharma Mar, Sanofi, Spectrum Pharmaceuticals, Takeda, Tiziana Pharma.