Lorlatinib in pretreated ALK- or ROS1-positive lung cancer and impact of TP53 co-mutations: results from the German early access program

Introduction: We report on the results of the German early access program (EAP) with the third-generation ALK- and ROS1-inhibitor lorlatinib. Patients and Methods: Patients with documented treatment failure of all approved ALK/ROS1-specific therapies or with resistance mutations not covered by approved inhibitors or leptomeningeal carcinomatosis were enrolled and analyzed. Results: In total, 52 patients were included [median age 57 years (range 32–81), 54% female, 62% never smokers, 98% adenocarcinoma]; 71% and 29% were ALK- and ROS1-positive, respectively. G1202R and G2032R resistance mutations prior to treatment with lorlatinib were observed in 10 of 26 evaluable patients (39%), 11 of 39 patients showed TP53 mutations (28%). Thirty-six patients (69%) had active brain metastases (BM) and nine (17%) leptomeningeal carcinomatosis when entering the EAP. Median number of prior specific TKIs was 3 (range 1–4). Median duration of treatment, progression-free survival (PFS), response rate and time to treatment failure were 10.4 months, 8.0 months, 54% and 13.0 months. Calculated 12-, 18- and 24-months survival rates were 65, 54 and 47%, overall survival since primary diagnosis (OS2) reached 79.6 months. TP53 mutations were associated with a substantially reduced PFS (3.7 versus 10.8 month, HR 3.3, p = 0.003) and were also identified as a strong prognostic biomarker (HR for OS2 3.0 p = 0.02). Neither prior treatments with second-generation TKIs nor BM had a significant influence on PFS and OS. Conclusions: Our data from real-life practice demonstrate the efficacy of lorlatinib in mostly heavily pretreated patients, providing a clinically meaningful option for patients with resistance mutations not covered by other targeted therapies and those with BM or leptomeningeal carcinomatosis.

BM at first diagnosis as well as during the course of disease. 6 Median progression-free survival (PFS) using the first-generation tyrosine kinase inhibitor (TKI) crizotinib ranges between 7.7 months in ALK-positive and 19.1 months in ROS1 patients, respectively. [7][8][9] Due to a low intracranial penetration rate, the brain represents the most prominent site of progression on crizotinib treatment. 5,[10][11][12] Consequently, first-line treatment with more potent second-generation TKIs has become the standard of care at least for ALK-positive patients. 13 However, the majority of patients inevitably relapses also using these newer drugs, revealing the medical need for sequential treatment options. 13,14 Acquired mutations in the ALK or ROS1 kinase domain represent a major molecular mechanism of resistance. Lorlatinib is a potent and selective third-generation, ATP-competitive oral ALK and ROS1 kinase inhibitor, especially designed to penetrate the blood-brain barrier and to overcome known ALK and ROS1 resistance mutations. Clinical trials including ALK-and ROS1-positive patients demonstrated the efficacy of lorlatinib in subsequent lines of therapy with high intracranial response rates (RRs). 15, 16 We here report on the results of the German early access program (EAP) of lorlatinib, providing data on mostly heavily pretreated patients with ALK-and ROS1alterated NSCLC from the daily routine.

Patients and methods
Patients with documented treatment failure of all approved ALK/ROS1-specific therapies and ⩾2 other approved systemic therapies for metastatic NSCLC could be enrolled into the EAP from April 2017 until May 2019. Patients with documented resistance mutations not covered by other inhibitors (e.g. G1202R for ALK and G2032R for ROS1) or leptomeningeal carcinomatosis (LMC) could receive lorlatinib even without having been treated with all approved lines of therapy. Resistance testing was not mandatory prior to enrolment. LMC was defined as the combination of multifocal neurologic signs, typical radiomorphologic findings in brain/spine magnetic resonance imaging (MRI) (e.g. diffuse leptomeningeal contrast enhancement), cytologic identification of malignant cells within the cerebrospinal fluid (CSF) and/or a CSF composition compatible with LMC (e.g. high protein concentration, low glucose concentration, lymphocytic pleocytosis). Detailed inclusion and exclusion criteria were as follows.

Inclusion criteria
1. Diagnosis: Evidence of histologically or cytologically confirmed diagnosis of metastatic NSCLC (Stage IV, AJCC v7.0) carrying an ALK or ROS1 rearrangement as determined by locally approved tests.

Disease Status Requirements:
Documented treatment failure (i.e. disease progression, symptom deterioration or intolerance to therapy) of all locally approved ALK/ROS1 inhibitor therapies and any other alternative approved systemic treatments for metastatic NSCLC had to be documented as a prerequisite.
For patients with ALK-positive NSCLC: •  10. Current use or anticipated need for drugs that are known strong CYP2C8 inhibitors, including their administration within 12 days prior to study entry (i.e. gemfibrozil). 11. Current use or anticipated need for drugs that are known P-gp substrates with a narrow therapeutic index, including their administration within 12 days prior to study entry (i.e. digoxin). 12. Breastfeeding female patients (including patients who intend to interrupt breastfeeding). 13. Known hypersensitivity to lorlatinib or any of its excipients.
Patients' baseline demographics, tumor-specific data and outcome were collected. As RECISTbased radiologic evaluation is not routinely performed in the real-world setting, estimation of tumor response and definition of disease progression were based on the clinician's estimates, which have nevertheless been shown to correlate reasonably well with RECIST assessments in several recent studies. 17,18 On-treatment imaging was performed according to national guidelines 19 and the respective local standard of care, respectively, using computed tomography scans (thorax, abdomen) and brain MRIs. Positron emission tomography scans were not routinely performed.
Survival endpoints were real-world PFS, time to treatment failure (TTF) and overall survival (OS). PFS was defined as the time in months from the first dose of lorlatinib to the first documented progression, either as radiologically confirmed progression or death, PFS2 as the time in months from the first dose of any subsequent therapy following lorlatinib until death from any cause, TTF as the time in months from the first dose of lorlatinib until loss of clinical efficacy as defined by the treating physician, OS1 as the time in months from the first dose of lorlatinib and OS2 since the date of primary diagnosis until death from any cause.
AEs were graded according to the CTCAE version 5.0. Approval was obtained from the Charité Universitätsmedizin Berlin ethics committee (approval number EA2/159/19).

Statistical analysis
Demographics and disease data were described and compared using the Pearson Chi 2 -test, Fisher's exact test or Mann-Whitney U test, according to the level of measurement. The Kaplan-Meier method was used to estimate median PFS, TTF and OS. p-values comparing survival curves were calculated with log-rank tests. Hazard ratios were calculated using univariate Cox-regression analysis. All analyses were performed using IBM SPSS statistics version 24 (IBM, Armonk, NY, USA). A p-value < 0.05 (two-tailed) was defined as statistically significant.

Molecular characterization and identification of potential resistance mechanisms
Next-generation sequencing (NGS)-based analyses of potential tyrosine kinase resistance mutations prior to therapy with lorlatinib were performed in  Figure 1A) and overall survival since treatment with lorlatinib (OS1, Figure 1B) for ALK and ROS1-positive patients.
26 patients (50.0%,  Figure 3A). By trend, survival differences were also observed for OS1 (10.9 versus 24.7 months, p = 0.24, Figure  3B). In general, TP53-mutations carried out a strongly negative prognosis throughout the entire course of disease with an OS2 of 42.2    Figure 3B) and overall survival since primary diagnosis (OS2, Figure 3C), each depending on the TP53 mutational status.

Adverse events
AEs were reported in 37 patients (71.2%); the median number per patient was two (range 1-8; 14 patients without reported AEs, information missing n = 1). All AEs are listed in Table 3. The most frequent clustered AEs concerned gastrointestinal disturbances (34 patients), followed by neurological and respiratory disorders (16 and 11 patients, see footnote below Table 3 for a detailed description of all neuro-psychiatric disorders). The most common isolated AEs were hypercholesterolemia (n = 17) and peripheral edema (n = 11). The majority of AEs were mild or moderate grade (G1 or G2, 59 events, 37 patients). G3 and four events occurred in eight (nine events) and two patients (two events), respectively, concerning laboratory abnormalities in lipid metabolism and pancreatic enzymes as well as edema, central nervous system effects and pneumonitis. Grading was not reported for 21 events. Therapy with lorlatinib was discontinued due to AEs in five patients, suffering from dyspnea (G3), anasarca (G3), an increase in serum creatinine [grade not reported (NR)] and pneumonitis (G4). Psychiatric disorders leading to treatment discontinuation concerned hallucinations and persecution mania (grade NR) in two ALK-positive patients with BM and LMC.

Discussion
Our data from real-world treatment demonstrates the efficacy of lorlatinib in mostly heavily pretreated patients with either ALK-or ROS1rearrangements. RR was 42.4% for ALK-positive patients (n = 14), comparable to a phase II study 16 and the results from the expanded access program in Asian countries and the US. 20 In contrast to these studies, RR was not influenced by the number of prior TKIs and was also substantially higher for ROS1-positive patients (84.6%) as compared with the respective phase I/II study. 15 The brain represents a frequent site of progression in ALK-and ROS1-positive NSCLC, affecting nearly 60% of patients after 3 years. 21 In the crizotinib-refractory setting, intracranial RR for alectinib, brigatinib and ceritinib ranged between 35 and 73%. 22 According to the presence versus absence of BM at baseline, lorlatinib showed rates of brain progression after 12 months of 22 versus 9% after crizotinib, and 23 versus 12% after ⩾1 second generation ALK-TKI. 23 Although BM have not been evaluated separately in our investigation, the reported RRs were higher in patients with BM (62.5% versus 35.7%) suggesting a high intracranial efficacy. In contrast to the published literature, the brain also represented the most frequent site of progression in patients suffering from BM when entering the EAP (75.0%), whereas no patient without BM progressed intracranially. Due to these uneven results, a clear brain-protecting effect of lorlatinib could not be identified from our dataset. LMC represents a difficult to treat manifestation of BM, mostly refractory to standard treatment and associated with a dismal prognosis. In this setting, case series suggest a high efficacy using effective targeted therapy even after treatment failure of preceding ALK TKIs. 24 The prevalence of LMC in NSCLC in general is estimated to range between 3% and 5% 25 and substantially increases in molecularly altered subgroups, benefiting from a longer survival due to targeted therapies. The 25% of patients with LMC in our study may represent an enriched share of patients even for this special population, 20 showing a promising RR of 77.8%.
The median PFS reached 8.0 months and was 7.1 and 11.0 months for ALK and ROS1 patients, consistent with phase I/II studies reporting a PFS of 7.3 and 8.5 months in the ALK and ROS1 subsets, respectively. 15,16 Treatment beyond progression was continued in almost half of the patients with PD (14 patients, 41.2%) and resulted in a meaningful increase in TTF of 13.0 months. Treatment continuation beyond PD given an ongoing clinical response nowadays represents an established option for patients with molecularly altered NSCLC and has been proven beneficial. Hence, retrospective data suggest that these patients might represent a subset with a more favorable prognosis in general. 26 In the era of molecularly driven therapies, offering sequential targeted treatment options in ALK and ROS1-positive NSCLC, re-characterization of the tumor via repeated biopsies has become an established procedure and was carried out in 50% of patients in the routine setting. G1202Rsolvent-front mutations are associated with clinical resistance to first and second-generation ALK TKIs but may be sensitive to lorlatinib. 27 G2032R-mutations represent a similar situation for ROS1-patients. The available molecular data from the present investigation confirm the clinical efficacy of lorlatinib for both solvent-front mutations, whereas PFS and TTF were clearly reduced with the evidence of compound mutations. Furthermore, TP53 mutations were associated with a markedly reduced PFS of 3.7 versus 10.8 months. As differences in OS since primary diagnosis depending on the TP53 status were even more pronounced (42.2 versus 88.9 months), these alterations might represent an intrinsic mode of TKI resistance. In this line, several investigations have identified TP53 mutations, either present at diagnosis or acquired at disease progression, as a predictive as well as prognostic biomarker in ALK-positive NSCLC. 28,29 To our knowledge, our study is the first report transferring the validity of these results also on a treatment with lorlatinib, indicating that TP53 mutations confer a negative impact on PFS and OS irrespective of the TKI.
Whether re-exposure to a frontline TKI after progression on a subsequent TKI may be beneficial remains an unresolved question. In our study, patients receiving another TKI after lorlatinib had a shorter PFS2 than those treated with chemotherapy. However, selection bias may represent a significant confounder, as patients with a subsequent TKI may have been judged not fit enough for chemotherapy and data on post-lorlatinib PS were not reported. A significant fraction of the so far identified lorlatinib-resistant compound mutations is not sensitive to earlier generation TKIs. [30][31][32] Therefore, a re-exposure without a carefully performed molecular analysis suggesting sensitivity to first-or second-generation TKIs has to be regarded cautiously. Thus, patients may benefit more from chemotherapy.
The reported rates of AEs in our study were substantially smaller than those in the respective clinical trials, probably attributable to a less rigorous reporting in the daily routine. However, as patients were not treated according to a dedicated study protocol, dose modifications may have been used preemptively to avoid (more severe) AEs. It is noteworthy, that two of five patients with AE-related treatment cessation were discontinued due to psychiatric disorders, effects that have not been reported for other ALK or ROS1-TKIs. Especially BM (e.g. next to the limbic system) and LMC might predispose to psychiatric AEs. In this connection, prior brain radiotherapy and steroids administered to most patients with symptomatic BM have been described as potential trigger factors. 33 Albeit the biologic connection with the administration of lorlatinib is uncertain, it would be highly desirable to identify patients at risk in advance, as these AEs have a substantial impact on the patients' quality of life and might require discontinuation of a highly effective treatment.
Due to its retrospective design, this study has some limitations. ALK fusion variants, proven to impact response and survival on TKIs, were not assessed routinely and thus were not evaluable. 34,35 On the other hand, TP53 status and molecular resistance mechanisms at the time of disease progression were assessed at all participating centers in laboratories and by NGS platforms certified by the quality management initiative of the German Society of Pathology (QuIP ® ), therefore the results can be considered homogenous and reliable. Patients were treated within the valid standard of care outside a clinical trial with varying imaging intervals, potentially biasing PFS, and imaging was not routinely performed using RECIST assessments. In this line intra-and extracranial responses were not assessed separately and are not reported from our cohort. Nonetheless and with special regard to these limitations, a growing body of evidence suggests that radiologic outcome generated from a real-life cohort may be comparable to a RECISTdefined study cohort. 17,18 An important limitation results from the special patient characteristics of nearly every EAP. Data from real-world practice demonstrate a substantial loss of patients in the transition from one line of therapy to the other. 36 The recently reported FLAURA trial comparing osimertinib with erlotinib in EGFR-mutant NSCLC indicated that 22% of patients with PD did not receive any second-line treatment and 46-49% had no third-line therapy. 37 Thus, patients included into an EAP represent a clearly positively selected population and immortal-time bias should be taken into consideration with regard to OS. Results must be regarded with caution and may not be translated to a general population.

Conclusion
Our data from a real-world setting confirm the effectiveness of lorlatinib in heavily pretreated ALK and ROS1-positive patients, in whom all treatment options are exhausted. Control of BM is promising, especially as those patients are at a high risk to develop BM during the course of their disease. Integration of the TP53 status into molecular testing strategies may provide a helpful tool in the management of disease, as patients