Treatment-Related Adverse Events of Combination EGFR Tyrosine Kinase Inhibitor and Immune Checkpoint Inhibitor in EGFR-Mutant Advanced Non-Small Cell Lung Cancer: A Systematic Review and Meta-Analysis

Simple Summary The use of combination epidermal growth factor tyrosine kinase inhibitor (EGFR-TKI) and immune checkpoint inhibitor (ICI) in EGFR-mutant, advanced non-small cell lung cancer (NSCLC) has raised concerns over the risk of overlapping toxicities. Although a higher proportion of interstitial lung diseases was reported with the combination of osimertinib and durvalumab, the current evidence on the treatment-related adverse events (trAEs) of EGFR-TKI and ICI remains limited to pneumonitis and the use of osimertinib. This systematic review and meta-analysis investigates whether combination EGFR-TKI and ICI increases the incidence of overall and organ-specific trAEs compared to TKI monotherapy. A higher proportion of high-grade organ-specific trAEs was observed in combination TKI and ICI including skin, gastrointestinal adverse events, and interstitial lung diseases. Further prospective studies are warranted to determine the optimal drug of choice of ICI and the best timing of initiation of ICI after failure to prior EGFR-TKI. Abstract (1) Background: We performed a meta-analysis to examine whether combined epidermal growth factor tyrosine kinase inhibitor (EGFR-TKI) and immune checkpoint inhibitor (ICI) increases treatment-related adverse events (trAEs) in advanced non-small cell lung cancer (NSCLC). (2) Methods: Articles from MEDLINE, EMBASE, and Cochrane databases were searched. Proportions and odds ratios (ORs) of the pooled incidence of overall and organ-specific trAEs in combination EGFR-TKI and ICI were compared to TKI monotherapy. (3) Results: Eight studies fulfilled our selection criteria. Any-grade organ-specific trAEs were more common in combination EGFR-TKI and ICI than TKI monotherapy (skin: OR = 1.19, p = 0.012; gastrointestinal tract: OR = 1.04, p = 0.790; ILD: OR = 1.28, p = 0.001). Grade ≥ 3 trAEs were also more frequent in combination treatment (skin: OR = 1.13, p = 0.082; gastrointestinal tract: OR = 1.13, p = 0.076; ILD: OR = 1.16, p = 0.003). (4) Conclusions: A higher proportion of grade ≥3 skin and gastrointestinal trAEs and ILDs was observed in combination TKI and ICI compared to TKI alone. Caution has to be taken when interpreting the results owing to the small number of studies included in this meta-analysis.


Introduction
Sensitizing somatic mutations of the epidermal growth factor receptor (EGFR) gene were found in about 10% of Caucasian and 50% of east Asian patients with non-small cell randomized controlled trials (non-RCTs), and retrospective studies without language restriction. The keywords were non-small cell lung cancer or non-small-cell lung cancer, EGFR, immunotherapy, and immune checkpoint inhibitors. The search date started from the inception of each database to 31 December 2021. Abstracts and presentations from conference proceedings including American Society of Clinical Oncology, World Conference on Lung Cancer, and European Society for Medical Oncology were also reviewed to identify unpublished studies. The clinicaltrials.gov database (https://www.clinicaltrials.gov, accessed on 31 December 2021) was also screened to identify clinical trials that had not been published.

Selection Criteria
RCTs, non-RCTs, single-arm studies, and retrospective studies that fulfilled the following criteria were included: (i) recruited patients aged 18 years and above with advanced NSCLC (stage IIIB-IV) who were previously treated with EGFR-TKI, (ii) evaluated the safety and tolerability of combination EGFR-TKI and ICI, (iii) administered EGFR-TKI and ICI regardless of sequence, either concurrently or sequentially, and (iv) reported incidence of trAEs that included any-grade and grade ≥ 3 events. Trials were excluded if they included patients who were TKI-naïve or used ICI in combination with chemotherapy or other regimens of ICI, were not available in full text, and were case reports.

Study Objectives
The primary objective was to evaluate the pooled incidence of overall and organspecific including gastrointestinal, pulmonary, and dermatological trAEs of combination EGFR-TKI and ICI. The severity of trAEs was recorded as grade 1-5 based on version 3 or 4 of the Common Terminology Criteria for Adverse Events (CTCAE) of the National Cancer Institute (Bethesda, MD, USA). The secondary objective was to compare the pooled incidence of trAEs between combination therapy and TKI monotherapy. Previous literature suggested that osimertinib may carry a higher risk of trAEs and ILD [34,35,[37][38][39]. As result, FLAURA [15] and AURA3 [45], the two registration trials on osimertinib as first-line treatment for patients with treatment-naive EGFR-mutant metastatic NSCLC and secondline treatment for patients with EGFR T790M mutation positive metastatic NSCLC after failure to prior TKI, as well as the osimertinib monotherapy arm of CAURAL study were selected as control for comparison of incidence of trAEs. The summary of osimertinib monotherapy studies is listed in Supplementary Table S1.

Data Extraction
Data extracted were recorded using Excel 2013 on a predesigned extraction list. For clinical trials, information extracted included: trial name, name of first author, year of publication, stage of NSCLC, EGFR-TKI treatment status, incidence of any-grade and grade 3 or above adverse events, and incidence of adverse events leading to drug discontinuation or death. For retrospective studies, patient characteristics, previous oncological treatment, trAE onset, current treatment details, and outcomes were recorded.

Assessment of Study Quality
The quality of RCTs was evaluated by the Cochrane Collaboration "risk of bias" tools [46], and that of non-RCTs was assessed by the Newcastle-Ottawa Scale (NOS) [47].

Outcome Measures and Statistical Analysis
The outcome measure was the proportion of trAE incidence. Freeman-Tukey double arcsine transformation was applied to transform proportions to continuous scale for metaanalysis. Odds ratios (ORs) were further generated to compare the pooled proportion of trAE incidence of combination therapy against TKI monotherapy. It has to be noted that one osimertinib monotherapy study was excluded from the pooled analysis of overall trAEs due to insufficient data available [48].
Statistical heterogeneity and inconsistency between the selected studies were assessed with the I 2 statistics and τ 2 statistics, respectively. If I 2 > 50% or p value < 0.1 for the τ 2 test, which indicated potential heterogeneity and inconsistency amongst the studies, a random effects model was used. Otherwise, the analysis was based on fixed effect models. We also performed sensitivity analysis on trAEs that had studies with at least moderate sample size ≥ 40 subjects reported in both monotherapy and combination groups.
Since patients' individual data were not reported, the most common organ-specific disease was chosen to represent the incidence of the overall organ-specific trAE. For example, in Arm 1 of TATTON trial (osimertinib + duvalumab 3 mg/kg), vomiting (70%) was the most common gastrointestinal AE compared to nausea (30%) and diarrhea (30%). Therefore, the incidence of vomiting was chosen to represent the gastrointestinal AE of that arm. Statistical significance was defined as p value < 0.05 (two-sided). All data analysis was conducted using the statistical software R (version 3.6.3 (29 February 2020); R Foundation, Vienna, Austria) and the meta and metafor package.

Study Characteristics of Clinical Trials
One RCT and two non-RCTs were selected, as summarized in Table 1. In total, 73 patients with stage IIIb/IV EGFR-mutant NSCLC were included. All were TKI-pretreated except for 1 patient in CheckMate 012 (NCT01454102) [70]. While CAURAL (NCT02454933) [35] was a phase III RCT, TATTON (NCT02143466) [34] and CheckMate 012 were phase I singlearm studies. Both TATTON and CAURAL evaluated the safety and tolerability of osimertinib plus durvalumab, while CheckMate 012 investigated the combination of erlotinib and nivolumab.
All studies described the incidence of trAEs reported in ≥10% of patients graded according to CTCAE version 4. While CheckMate 012 reported organ-specific adverse events, other studies did not. As mentioned previously in the section "Data analysis", if a study does not report organ-specific AE, the most common organ-specific disease was chosen to represent it. Only gastrointestinal and skin AE as well as interstitial lung diseases (ILD) had data reported by all studies. Hepatotoxicity was reported by CheckMate 012 and CAURAL but not in TATTON. Upper respiratory diseases were reported by CUARAL but not in CheckMate 012 and TATTON. Endocrinopathies were reported by CheckMate 012 but not in CAURAL and TATTON. Details on the incidence and nature of trAEs reported by each study can be found in Supplementary Table S3. The risk of bias of RCTs and the quality assessment of non-RCTs were also shown (Supplementary Tables S4 and S5). All studies described the incidence of trAEs reported in ≥10% of patients graded according to CTCAE version 4. While CheckMate 012 reported organ-specific adverse events, other studies did not. As mentioned previously in the section "Data analysis", if a study does not report organ-specific AE, the most common organ-specific disease was chosen to represent it. Only gastrointestinal and skin AE as well as interstitial lung diseases (ILD) had data reported by all studies. Hepatotoxicity was reported by CheckMate 012 and CAURAL but not in TATTON. Upper respiratory diseases were reported by CUARAL but not in CheckMate 012 and TATTON. Endocrinopathies were reported by CheckMate 012 but not in CAURAL and TATTON. Details on the incidence and nature of trAEs reported by each study can be found in Supplementary Table S3. The risk of bias of RCTs and the quality assessment of non-RCTs were also shown (Supplementary Tables  S4 and S5).

Study Characteristics of Retrospective Studies
All five retrospective studies reported ILD or pneumonitis in advanced, TKI pretreated NSCLC patients who have received sequential TKI and ICI therapy [36][37][38][39]71]. One study was based on the US Food and Drug Administration Adverse Event Reporting System (FAERS) database analysis [36], three were single-institution studies [37][38][39], and one was a multi-institutional study [71]. The summary of these studies was shown in Table 2.

Study Characteristics of Retrospective Studies
All five retrospective studies reported ILD or pneumonitis in advanced, TKI pretreated NSCLC patients who have received sequential TKI and ICI therapy [36][37][38][39]71]. One study was based on the US Food and Drug Administration Adverse Event Reporting System (FAERS) database analysis [36], three were single-institution studies [37][38][39], and one was a multi-institutional study [71]. The summary of these studies was shown in Table 2.

Nature of trAEs Reported in Retrospective Studies
Four retrospective studies described the incidence of ILD in sequential or concurrent administration TKI and ICI [36][37][38]71]. Only one study reported severe immune-related adverse events other than ILD, including colitis and hepatitis [37]. Oshima et al. revealed a higher proportion of ILD for nivolumab in combination with EGFR TKI compared to treatment with either drug alone [36]. Uchida et al. found administration of first or second-generation TKI immediately after nivolumab did not increase the risk of ILD, but ILD was more frequent if osimertinib was used immediately after ICI [39]. Kotake et al. also highlighted an increased risk of ILD in the use of nivolumab immediately followed by osimertinib [38]. In this study, the median time to the onset of ILD from osimertinib initiation was 6 weeks (range 4-12) with a median interval of 2 weeks (range 1-4). In addition, PD-L1 blockade followed by osimertinib was associated with an increased risk of severe trAEs [37]. However, this association appeared to be specific to osimertinib, as no severe trAES occurred with the administration of other TKIs. In general, trAEs occurred at a median onset of 20 days (range 14-167 days) after osimertinib. In a recent retrospective study investigating the efficacy and safety of the use of ICI monotherapy after TKI, no ILD was reported. The ICI monotherapy included nivolumab, pembrolizumab, or atezolizumab and the median period between termination of TKI and initiation of ICI was 4.5 months (0.03-22.8) [71].

Incidence of trAE Leading to Death and Drug Discontinuation
Only TATTON reported death in a patient after receiving osimertinib plus 3 mg/kg durvalumab due to pneumonia. However, this was not considered related to treatment. No trAE led to death in other combination therapy studies. In trials of osimertinib monotherapy, a pooled proportion of 7% (95% CI: 0-14) of trAEs led to death. It was also revealed that trAE led to discontinuation of ICI in 14% (95% CI: 7-22) of cases and TKI in 12% (95% CI: 2-21) of cases. In osimertinib monotherapy, 7% (95% CI: 0-14) of trAE led to the drug discontinuation.

Discussion
Immune checkpoint inhibitors, especially anti-PD-(L)1 antibodies, have revolutionized the treatment paradigm of advanced NSCLC, especially those without targetable driver mutations. However, the combinational use of EGFR-TKI and ICI, regardless of whether they are used concomitantly or sequentially, has raised serious concerns over undesirable toxicities. To the best of our knowledge, our study is the first meta-analysis and systematic review of trAEs in combination EGFR-TKI and ICI in advanced EGFR-mutant NSCLC. The most significant finding of the current study was that the pooled incidences of grade ≥ 3 skin and gastrointestinal trAEs and ILD were significantly higher in combination therapy than osimertinib monotherapy. This finding is consistent with previous reports of combining PD-L1 inhibitors and EGFR-TKIs. The recent FAERS database study revealed 25.7% (18 of 70) of patients treated with combination EGFR-TKI and nivolumab experienced ILD of any-grade [36], which echoed our findings (16%, Table 3). Although the incidence of ILD reported here was lower, the actual rate of adverse events may be higher because our analysis considered only the most common organ-specific reported trAE as the overall organ-specific trAE given the unavailability of patients' individual data from reported studies.
It should be noted that retrospective studies were included in this review to complement both the quantitative meta-analysis and the qualitative discussion of the nature of trAEs. The current evidence of retrospective studies shows a higher risk of ILD and other severe trAEs in the use of osimertinib immediately (within 180 days) after nivolumb [38,39] or other PD-L1 antibodies [37]. The association appeared to be specific for osimertinib, as administration of first or second-generation TKIs immediately after nivolumab did not increase the risk of ILD [39], and no trAEs were observed when PD-L1 blockers were followed by other EGFR-TKIs [37,71]. The reason why ILD development is different between first/second-generation TKI and osimertinib remains unclear. The activation of T-cell effects by ICI may be upregulated to cause ILD in osimertinib but not in first/second-generation TKI [39].
Several case reports also revealed a higher risk of ILD in the sequential use of nivolumab followed by osimertinib within a short interval. The onset of ILD was reported to be on the 63rd day in a 37-day interval [49] and the 50th day in a 46-day interval [50]. Both cases were treated with steroids and ILD remained in remission. It is noteworthy that in a case of osimertinib-induced ILD, osimertinib re-challenge 8 months after the last initiation of nivolumab led to remarkable tumor shrinkage without evidence of relapse of osimertinib-induced ILD [51].
The mechanism of the synergistic toxicity of these two therapies remains undeciphered. It may be related to prolonged receptor occupancy of PD-L1 antibodies [37,50], higher levels of proinflammatory cytokines observed in the combination therapy [43], and enhanced T-cell mediated killing of tumor cells by cytokine-induced MHC class-I presentation [42]. Nivolumab can occupy PD-1 receptors on T-cells for approximately two months, which may contribute to delayed trAEs. A recent study of liquid chromatography-mass spectrometry analyses revealed that nivolumab concentrations lower than the trough concentrations during nivolumab and osimertinib treatment were high enough to induce ILD [50].
Our study also revealed increased risks of high-grade skin and gastrointestinal AEs in combination TKI and ICI versus TKI alone, albeit a weak association. Other organ-specific toxicities such as hepatotoxicity were not included into this meta-analysis due to limited data available. Increased levels of alanine aminotransferase up to 40% were reported in patients receiving a combination of gefitinib and durvalumab [72].
Our analysis has some limitations. First, our findings mainly focused on trials of combination osimertinib plus durvalumab. Additional data will be needed to clarify the relative risks of other TKIs following PD-L1 blockade. The exclusion of case reports may result in selection bias, even though the ability to draw statistically meaningful conclusions from case reports is extremely limited. Further, there were only a few randomized and prospective clinical trials studying the use of ICI and EGFR-TKI for advanced NSCLC. Thus, the number of studies included in this meta-analysis is small. The limited patient size also hinders us from performing clinically meaningful statistical analysis to further compare the risk of AEs between concurrent and sequential treatment of the two agents. In the sensitivity analysis which considered studies with moderate sample size ≥ 40, the conclusion remained that combination therapy generated higher risk of ILD when compared with monotherapy. While ILD was reported to be higher in the Japanese population [34], subgroup analysis was not possible due to the lack of reporting on patient demographics. In addition, although ILD has been reported to be more frequent in patients treated with osimertinib, it is difficult to examine the difference in the incidence of ILD among every TKIs. While most studies that reported ILD involved combination ICI and osimertinib [34,35,38], only one investigated erlotinib [70] while the others did not state the TKI that caused ILD [36,37,39,71]. Larger prospective studies are warranted to investigate the incidence of trAEs between concurrent and sequential treatment of ICI and TKI, and to stratify the risks of ILD according to different ethnicities and TKIs. Furthermore, we were unable to investigate and compare the incidences of other clinically important trAEs including endocrinopathies, hepatotoxicities, nephrotoxicities, and neurotoxicities since they were rarely and less systematically reported in every selected study. Moreover, toxicities in some patients might have been attributed to other causes such as progressive disease of their underlying NSCLC, which would underestimate the reported treatment-related toxicity and were inherent limitations of retrospective analyses. Lastly, our study results cannot serve as a recommendation guideline on how and when to administer TKIs and ICI judiciously, as these agents could be safely administered in some patients in the clinical trials included in this meta-analysis.

Conclusions
Our meta-analysis demonstrates, in general, an increased overall incidence of trAE and organ-specific trAE in combination therapy with EGFR-TKI and ICI. Future prospective RCTs are warranted to determine the most optimal drug of choice of ICI and the best timing of initiation of ICI after failure to prior EGFR-TKI.  Table S1: summary of osimertinib monotherapy studies; Table S2: ongoing clinical trials of immune checkpoint inhibitors (ICI) in advanced, metastatic non-small cell lung cancer; Table S3: incidence and nature of trAEs in selected clinical studies; Table S4: the Cochrane Collaboration's tool for assessing risks of bias of randomized-controlled trials; Table S5: newcastle-Ottawa Scale (NOS) for quality assessment of non-randomized-controlled trials.