Background: Atypical EGFR mutations occur in 10%-30% of non-small-cell lung cancer (NSCLC) patients with EGFR mutations and their sensitivity to classical epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors (TKI) is highly heterogeneous. Patients harboring one group of uncommon, recurrent EGFR mutations (G719X, S768I, L861Q) respond to EGFR-TKI. Exon 20 insertions are mostly insensitive to EGFR-TKI but display sensitivity to exon 20 inhibitors. Clinical outcome data of patients with very rare point and compound mutations upon systemic treatments are still sparse to date. Patients and methods: In this retrospective, multicenter study of the national Network Genomic Medicine (nNGM) in Germany, 856 NSCLC cases with atypical EGFR mutations including co-occurring mutations were reported from 12 centers. Clinical follow-up data after treatment with different EGFR-TKIs, chemotherapy and immune checkpoint inhibitors were available from 260 patients. Response to treatment was analyzed in three major groups: (i) uncommon mutations (G719X, S7681, L861Q and combinations), (ii) exon 20 insertions and (iii) very rare EGFR mutations (very rare single point mutations, compound mutations, exon 18 deletions, exon 19 insertions). Results: Our study comprises the largest thus far reported real-world cohort of very rare EGFR single point and compound mutations treated with different systemic treatments. We validated higher efficacy of EGFR-TKI in comparison to chemotherapy in group 1 (uncommon), while most exon 20 insertions (group 2) were not EGFR-TKI responsive. In addition, we found TKI sensitivity of very rare point mutations (group 3) and of complex EGFR mutations containing exon 19 deletions or L858R mutations independent of the combination partner. Notably, treatment responses in group 3 (very rare) were highly heterogeneous. Co-occurring TP53 mutations exerted a non-significant trend for a detrimental effect on outcome in EGFR-TKI-treated patients in groups 2 and 3 but not in group 1. Conclusions: Based on our findings, we propose a novel nNGM classification of atypical EGFR mutations.