Sequential treatment with afatinib and osimertinib in patients with EGFR mutation-positive non-small-cell lung cancer: an observational study

Over the past decade, treatment of EGFR mutation-positive non-small-cell lung cancer (NSCLC) has been revolutionized with the development of EGFR tyrosine kinase inhibitors (TKIs). These agents are associated with improved clinical outcomes and improved tolerability, compared with the previous standard of care, platinum-doublet chemotherapy. Three generations of EGFR TKIs are approved for the first-line treatment of EGFR mutation-positive NSCLC, based on positive Phase III trials: the first-generation reversible TKIs, erlotinib and gefitinib; the second-generation irreversible ErbB family blocker, afatinib; and the third-generation EGFR-wild-type sparing, irreversible EGFR/T790M inhibitor, osimertinib [1].

Recent head-to-head trials have demonstrated that second- and third-generation EGFR TKIs are preferable to first-generation EGFR TKIs in patients with EGFR mutation-positive NSCLC. In the LUX-Lung 7 trial, afatinib conferred significant improvements in progression-free survival (PFS; hazard ratio [HR]: 0.73, 95% CI: 0.57–0.95; p = 0.017) and time-to-treatment failure (TTF; HR: 0.73, 95% CI: 0.58–0.92; p = 0.0073) versus gefitinib with a trend toward improved overall survival (OS; HR: 0.86, 95% CI: 0.66–1.12; p = 0.2580) in patients with common EGFR mutations (Del19/L858R) [2,3]. In a similar patient population (but excluding patients with brain metastases), the second-generation TKI, dacomitinib significantly improved PFS (HR: 0.59, 95% CI: 0.47–0.74; p < 0.0001) and OS (HR: 0.76, 95% CI: 0.58–0.99; p = 0.044) versus gefitinib in the ARCHER-1050 trial [4,5]. In the recent FLAURA trial, osimertinib significantly improved PFS versus gefitinib/erlotinib (HR: 0.46, 95% CI: 0.37–0.57; p < 0.001) [6]. OS data are currently immature.

Regardless of which first-line EGFR TKI is chosen, acquired resistance to therapy is inevitable. Therefore, a key consideration when assessing therapeutic choices is the availability of subsequent treatment options following disease progression. The predominant molecular resistance mechanism to gefitinib, erlotinib and afatinib is the emergence of the T790M mutation in exon 20 of EGFR, which is present in approximately 50–70% of tumors at the time of acquired resistance [7–10]. Osimertinib was originally developed to target T790M-positive tumors, and is approved in this setting based on data from the AURA trials (median PFS of 9.6–12.3 months) [9,11,12]. In contrast to first- and second-generation TKIs, targeted therapy options following first-line osimertinib remain limited, due to heterogeneity of resistance mechanisms that are generally not fully understood or currently druggable [13,14]. As such, in most patients progressing on osimertinib in everyday clinical practice chemotherapy is the only choice for second-line treatment. There is an argument, therefore, for treating patients with sequential EGFR TKIs, with osimertinib reserved as a second-line option to maximize treatment time on targeted drugs and defer the need for more toxic chemotherapy regimens.

Few data are available that have assessed the cumulative benefit of sequential EGFR TKIs in patients with EGFR mutation-positive NSCLC. In the current observational, global, multicenter study, we have assessed total treatment duration of sequential afatinib and osimertinib in patients with EGFR mutation-positive NSCLC with T790M-aquired resistance who were treated in ‘real-world’ clinical practice.

Patients & methods

Study design

GioTag was a global, observational study conducted across ten countries (Austria, Canada, Israel, Italy, Japan, Singapore, Slovenia, Spain, Taiwan and the USA; NCT03370770). Medical and electronic health records of consecutive patients treated in a real-world practice who met the following criteria were retrospectively reviewed between 28 December 2017 and 31 May 2018: aged ≥18 years with EGFR-mutated (Del19/L858R) TKI-naive advanced NSCLC who were treated first-line with afatinib, developed the T790M mutation and received second-line osimertinib treatment, and provided written informed consent where required. Patients could have received osimertinib as part of a compassionate use or expanded access program. Patients were excluded if they had received any other first- or second-line treatments, had active brain metastases at the start of either afatinib or osimertinib therapy, or had been treated within a clinical trial. Inclusion was restricted to patients who initiated osimertinib treatment ≥10 months prior to enrollment to avoid early censoring and ensure mature data. A maximum of 15 patients were enrolled per site. Data were sourced either from sites directly approached by the sponsor (38%) or from electronic health records supplied by Cardinal Health (OH, USA). For quality assurance of the documented patient observations, source data verification was performed on approximately 30% of included patients.

Outcomes & assessments

The primary outcome was time on treatment, defined as the time from the first dose of afatinib to that of the last dose of osimertinib or death. The secondary outcome was the type and proportion of acquired resistance mechanisms after osimertinib treatment. Clinical subgroups were defined based on the characteristics (ethnicity, EGFR mutation, presence of brain metastases and Eastern Cooperative Oncology Group performance status [ECOG PS]) reported at baseline. Treatment interruptions and dose reductions for afatinib and osimertinib were also documented. Sites were requested to provide information on mutational status after osimertinib failure, if available. EGFR mutation and T790M status were determined as per local methodology and practice.

Statistical analysis

The median time on treatment was predicted to be approximately 24 months, assuming 13–14 months PFS with afatinib and 10–13 months with osimertinib [2,9,11,15]. A sample size of 171 patients would give an 80% likelihood to observe a 90% CI of median time on treatment ≤10 months; assuming 10% of observations would be censored, a sample size of 190 patients was determined.

Time on treatment and OS were estimated using the Kaplan–Meier method; for patients still on treatment, time on treatment was censored at the date of data collection. Baseline conditions and demographics, as well as analyses of dose modifications of afatinib and osimertinib were descriptive.

Results

Patients

A total of 221 patients were assessed for eligibility and 17 were excluded for protocol violations (Supplementary Figure 1). Baseline characteristics of the 204 patients included in the analysis are shown in Table 1. Patients had diverse ethnicity with 120 (58.8%) Caucasians, 50 (24.5%) Asians and 18 (8.8%) African–American patients; patient enrollment by country is shown in Supplementary Table 1. At the start of afatinib treatment, ECOG PS was 0, 1 or ≥2 in 43 (21.2%), 110 (54.2%) and 31 (15.3%) patients, respectively. A total of 135 patients (75%) maintained or even improved ECOG PS during afatinib treatment, and 43 (21.1%), 91 (44.6%) and 63 (30.9%) patients had ECOG PS 0, 1 or ≥2 at the start of osimertinib treatment. The proportion of patients with higher ECOG PS was greater at the start of osimertinib treatment. At the start of afatinib treatment, 150 (73.5%) patients had a Del19 mutation and 53 (26.0%) had the L858R mutation. One patient had both Del19 and L858R. At the start of osimertinib treatment, all patients had a documented T790M mutation. Baseline characteristics according to ethnicity are shown in Supplementary Table 2. Characteristics were generally consistent regardless of ethnicity, although the non-Asian subgroup (Caucasians and African–Americans) included fewer female patients (47.1 vs 66.0%) and more Del19-positive patients (78.3 vs 62.0%) than the Asian subgroup. The Asian subgroup included patients of lower bodyweight than the non-Asian subgroup (median weight 61.2 vs 74.4 kg).

Of 204 patients, 169 (83.7%) and 200 (98.0%) patients received starting doses of 40 mg/day afatinib and 80 mg/day osimertinib, respectively (Table 1). The predominant reason for discontinuation of afatinib was progressive disease (n = 190, 93.1%; Supplementary Figure 1). At the time of database lock (29 June 2018), 106 (52.0%) patients had discontinued osimertinib, with progressive disease being the most common reason for discontinuation (n = 98, 92.5%); 98 patients were continuing on osimertinib. Half of all patients (50.5%) started afatinib therapy before 2016.

Time on treatment

After a median follow-up of 28.2 months (range: 14.0–96.8 months), the median time on treatment for sequential afatinib and osimertinib was 27.6 months (90% CI: 25.9–31.3; Figure 1). Median time on afatinib was 11.9 months (90% CI: 10.9–12.2) and median time on osimertinib was 14.3 months (90% CI: 12.8–15.9). The median time between stopping afatinib and starting osimertinib was 16 days (interquartile range: 6–33).

Overall time on treatment, and time on afatinib and osimertinib treatment, according to patient subgroups (categorized by ethnicity, age, EGFR mutation type, presence of brain metastases and ECOG PS) are shown in Table 2 and Figure 2. Clinical benefit of sequential afatinib and osimertinib was seen across patient subgroups, with a prolonged median time on treatment reported for Asian patients (overall: 46.7 months; afatinib: 14.0 months; osimertinib: 19.6 months). There was no difference between Caucasian and African–American patients, with 27.6 months for both (Supplementary Table 2). Overall time on treatment (median 30.3 vs 19.1 months), as well as time on afatinib (median 12.6 vs 10.0 months) and time on osimertinib treatment (median 15.0 vs 8.3 months), were longer in patients with Del19 versus L858R mutations. Median time on treatment was also longer in patients with ECOG PS of 0/1 versus ≥2 (overall: 31.3 vs 22.2 months; afatinib: 12.0 vs 10.4 months; osimertinib: 15.9 vs 9.4 months). Patient age had no effect on median time on treatment (median 27.6 months for patients aged <65 and those aged ≥65 years). Overall time on treatment was numerically longer in patients without baseline brain metastases versus those with brain metastases (median 28.4 vs 19.4 months; afatinib 11.9 vs 10.4 months and osimertinib 14.3 vs 10.2 months). Of 183 patients with no brain metastasis at baseline, 12 (6.6%) developed brain metastasis while on afatinib. Of 21 patients with brain metastases at the start of afatinib therapy, 8 (38.1%) were reported to have no brain metastasis at the start of osimertinib therapy.

Overall survival

At the time of analysis, 63 patients had died (30.9%). Therefore, OS analysis was limited to landmarks of 2 and 2.5 years where maturity was 50.0 and 37.1%, respectively. The 2-year OS rate from start of afatinib was 78.9% (Figure 3). The 2.5-year OS rate was 68.8%. Restricting the analysis to patients with ECOG PS of 0/1, the 2-year landmark was 83.9% (Supplementary Figure 2).

Mutational status after osimertinib discontinuation

Of the 106 patients who had discontinued osimertinib, mutational status data were available only for 39. Of these 39 patients, 27 (69.2%) had retained both the T790M and EGFR sensitizing mutations, eight (20.5%) had lost the T790M mutation but retained the EGFR sensitizing mutation, and four (10.3%) patients had lost the EGFR sensitizing mutation but retained the T790M mutation. No genetic alterations other than the common EGFR mutations (Del19/L858R) and/or T790M were reported. No information on the methodology used was collected.

Discussion

This observational, global, multicenter study is the first to evaluate outcomes of patients who received first-line afatinib followed by osimertinib. We observed sustained clinical benefit with this regimen in a real-world clinical practice setting in patients with EGFR mutation-positive NSCLC, with a median time on treatment of 27.6 months reported for this broad patient population that also included patients who have not been well represented in prior studies, such as those with ECOG PS ≥2 (n = 31, 15.3%). Importantly, this clinical benefit was consistent across all patient subgroups, with particularly encouraging results seen for those with Del19-positive disease (median time on treatment 30.3 months) and Asian patients (median time on treatment 46.7 months). These findings suggest that sequential afatinib and osimertinib is an effective treatment option in patients with EGFR mutation-positive NSCLC who acquire T790M, and may be particularly attractive for Del19-positive disease, given that the likelihood of acquiring T790M is approximately 1.5-fold higher than for L858R-positive disease [9,10,16–18]. Around 75% of patients with Del19-positive disease will acquire T790M resistance, and afatinib as upfront therapy has shown a significantly prolonged OS versus chemotherapy in the Del19-positive patient subgroup [19].

This study complements a recent exploratory analysis of LUX-Lung 7, and the LUX-Lung 3 and 6 clinical studies (which compared first-line afatinib with platinum-doublet chemotherapy) [20]. In 37 patients who received osimertinib following afatinib, median duration of treatment on osimertinib was 20.2 months (95% CI: 12.8–31.5) After 4.7 years, median OS was not reached and the 3-year OS rate was approximately 90%. In addition, in a recent single-center study (n = 27), the median time on treatment on sequential afatinib and osimertinib was 25.0 months (95% CI: 20.0–33.0) [10]. Another recent observational study also demonstrated prolonged duration of treatment and survival (67 months) in patients receiving a third-generation EGFR TKI following a first- or second-generation EGFR TKI [21]. While OS data in the current study are immature, and the retrospective nature of the study does not allow for a control arm, the estimated 2-year and 2.5-year OS of 78.9 and 68.8% are encouraging, particularly noting the possible under-representation of patients who experienced long-term benefit from either drug due to study timelines. While there are little prospective data on the benefits of sequential TKI therapy, prolonged OS has been observed with second-generation TKIs in first-line clinical trial settings. Median OS with afatinib in LUX-Lung 3, with very little subsequent osimertinib use, was 31.6 months [19]. Median OS with dacomitinib in ARCHER-1050 (which excluded patients with brain metastases) was 34.1 months in the intention-to-treat population and 36.7 months in patients who received subsequent osimertinib [5].

A recent analysis concluded that prospective randomized controlled trials in oncology tend to show more effectiveness than seen in the real-world setting, especially if the randomized controlled trials have used surrogate markers such as PFS as end points and if the trial population is directly compared with the overall population, which includes patients not typically enrolled in clinical trials [22]. The prolonged treatment duration with first-line afatinib followed by osimertinib across a broad range of patient subgroups included in this study is therefore encouraging. Indeed, this is consistent with other real-world studies that have demonstrated similar, or better, real-world outcomes with afatinib than observed in clinical trials [23,24]. The median time on osimertinib treatment in this study is longer than the median duration of osimertinib treatment (8.1 months) and median PFS (10.1 months) seen in the AURA-3 clinical trial [11]. These differences might, at least in part, be driven by prolonged treatment beyond RECIST progression in everyday clinical practice. However, the prolonged duration of therapy with osimertinib after afatinib may reflect the observation that afatinib-treated NSCLC cells can show a very high allelic frequency of T790M at the point of acquired resistance (Kohsaka et al. submitted), suggesting that afatinib-resistant tumors are particularly homogenous at the cellular level and therefore may be particularly sensitive to a T790M-directed TKI.

As a real world-based, observational study, this analysis provides data on a broader collective of patients than typically assessed in previous clinical trials of afatinib or osimertinib. For example, 18 (8.8%) patients were African–American, a demographic that was scarcely represented in the LUX-Lung or other EGFR-TKI clinical trials, and 31 (15.3%) had an ECOG PS of ≥2, making our results particularly encouraging and applicable to everyday clinical practice. Also, the frequency of patients with Del19-positive tumors (73.5%) was higher than in the LUX-Lung trials of first-line afatinib (49–58%) but similar to that observed in the AURA-3 trial of second-line osimertinib (68%) [2,11,15,25]; probably reflecting the higher frequency of T790M acquired resistance in Del19-positive than L858R-positive tumors [16,17]. This observation may also indicate preferential use of afatinib in Del19-positive patients, due to previous OS data [19].

In addition to the particularly notable benefit seen in Asian patients and patients with Del19-positive disease, sequential afatinib and osimertinib conferred prolonged clinical benefit across the range of patient subgroups assessed in this study. Patients with brain metastases is an important subgroup, because CNS involvement is a common feature of EGFR mutation-positive NSCLC [26]. We found a clear time on treatment benefit with afatinib and osimertinib regardless of the presence of brain metastases, supporting previous results indicating that afatinib and osimertinib are effective options in this subgroup [27,28]. Of note, although brain imaging was not mandatory and patients with symptomatic brain metastases were excluded, only 12 of 183 (6.6%) patients developed brain metastasis during afatinib therapy, possibly reflecting the ability of afatinib to penetrate the blood–brain barrier [29–31]. Compared with Asian and Caucasian patients, little is known about outcomes in African–American patients with EGFR mutation-positive NSCLC, with few such patients enrolled in previous clinical trials of EGFR TKIs. While numbers are small (n = 18), the median time on treatment for African–American patients here was consistent with that seen for the overall population and for Caucasian patients. As might be expected, median time on treatment was significantly shorter for patients with ECOG PS ≥2 versus 0/1. However, it should be noted that in this real-world clinical practice population, patients with PS ≥2, who are rarely included in clinical trials, derived clinical benefit from sequential afatinib and osimertinib treatment, with a median time on treatment of 22.2 months. Consistent with previous studies [32,33], age did not appear to have a negative impact on benefit derived from EGFR TKI therapy; overall median time on treatment was 27.6 months in patients ≥65 years old.

The acquisition of T790M is the main mechanism of acquired resistance to afatinib [9,10,18] and osimertinib is a recommended standard second-line therapy for these patients [34]. In contrast, few data are available regarding resistance mechanisms following first-line osimertinib. Resistance mechanisms to later-line osimertinib appear to be heterogeneous. The most predominant mechanism identified to date is the acquisition of the tertiary EGFR mutation, C797S, which is present in approximately 11% and 22–40% of cases following first- or later-line administration of osimertinib [35,36]. Other putative resistance mechanisms, such as amplification (e.g., MET, KRAS) or mutation (e.g., MEK1, KRAS, PIK3CA, HER2, BRAF) of other genes, or histological transformation, have been identified but at lower frequency [35–37]. In the present study, only a minority of patients underwent molecular testing following failure of osimertinib, most likely reflecting the lack of available validated targeted treatment options in this setting in routine clinical practice. Moreover, the testing did not identify any druggable alterations. Given the lack of targeted treatment options following osimertinib failure, chemotherapy is the most prevalent treatment option. Indeed, in FLAURA, of the 59% of patients who received subsequent therapy following osimertinib, 56% received chemotherapy and 35% received an off-label EGFR-TKI-based regimen [38].

This study supports the feasibility of a sequencing strategy of afatinib followed by osimertinib in patients with EGFR mutation-positive NSCLC who acquire T790M. The observed turnaround time between discontinuation of afatinib and initiation of osimertinib (median 16 days) indicates that rapid T790M testing in the real-world clinical practice is feasible and that many physicians continue first-line EGFR-TKI treatment until test results are available. Given that the frequency of uptake of subsequent therapy following first-line afatinib is high (71–80% of patients in LUX-Lung 3, 6 and 7) [2,20], together with the high acquisition rate of T790M in patients receiving afatinib, it is likely that many patients with EGFR mutation-positive NSCLC would be eligible, and fit enough, to receive such a regimen. However, on the other hand, first-line use of osimertinib is supported by highly encouraging efficacy observed in FLAURA [6]. Reserving osimertinib for second-line use could potentially result in some patients (i.e., those whose tumors accrue T790M-independent mechanisms of resistance) never having the opportunity to receive it at all. Furthermore, first-line use of osimertinib is supported by its favorable tolerability profile, with a lower rate of grade ≥3 adverse events than typically observed with first- and second-generation TKIs [2,4,6]. Conversely, a treatment strategy involving sequential EGFR TKIs may effectively prolong the overall chemotherapy-free treatment period. Overall, therefore, while there is strong evidence for the use of first-line osimertinib, there remains an argument for considering sequential afatinib and osimertinib, particularly in patients with Del19-positive tumors. The relative benefits and shortcomings of the two strategies remain open to debate at this time and prospective clinical trials are required with OS and quality of life as primary end points.

The main limitations of this study were its retrospective nature and potential for selection bias. Methodological efforts were undertaken to minimize the potential for selection bias, such as including only consecutive patients who fulfilled all of the inclusion criteria and limiting enrollment to a maximum of 15 patients per site. Nevertheless, patients who died on afatinib or were unfit or unwilling for second-line therapy were excluded by the study design; this exclusion of patients who died on first-line afatinib potentially introduced an immortal time bias. However, in the LUX-Lung trials, only approximately 6% of patients died on first-line afatinib [15,25]. On the other hand, owing to the study timelines (patients must have started second-line osimertinib at least 10 months prior to data collection) and dates of drug approvals and availability (afatinib was approved/reimbursed in participating countries between the second half of 2013 and end of 2014), patients who derived long-term benefit from first-line afatinib (10–12% in the LUX-Lung trials) had little chance to be enrolled in the study and may therefore have been under-represented. The other main limitation of the study was a lack of a comparator arm, which limits interpretation of the results. The 10-month cutoff for starting osimertinib prior to data entry did not introduce selection bias because there was no specified lower limit on the duration of osimertinib treatment; patients did not have to receive >10 months of osimertinib treatment to be included in the study; instead, it was specified as an attempt to maximize maturity of data by reducing risk of early censoring.

Conclusion

Our results demonstrate the feasibility of sequential afatinib and osimertinib therapy in patients with EGFR mutation-positive NSCLC who acquire T790M. With median time on treatment of 27.6 months in a real-world setting and notably prolonged treatment duration in some patient subgroups, the regimen appears to represent a viable and potentially effective regimen. While prospective studies are needed to fully determine the optimum treatment strategy for patients with EGFR mutation-positive NSCLC, this sequential treatment might offer sustained clinical benefit to a substantial number of patients, while avoiding chemotherapy, and warrants consideration.