EGFR mutation testing has become an essential determination to decide treatment options for NSCLC. The mutation analysis is often conducted in samples with low percentage of tumour cells from primary tumour biopsies. There is very little evidence that samples from metastatic tissues are suitable for EGFR testing. We had evaluated the frequency of EGFR mutations with three highly sensitive PCR techniques in formalin-fixed, paraffin-embedded samples of 143 NSCLC patients with central nervous system (CNS) metastases. 32 corresponding primary tumours were also examined. We used PCR followed by DNA fragments length analysis (FLA), ASP-PCR and PNA-LNA PCR clamp techniques. We found 9 (6.29 %) EGFR gene mutations in CNS samples: 3 (2.1 %) in exon 19 and 6 (4.2 %) in exon 21. The full concordance between CNS metastases and primary tumour samples was observed. PCR followed by DNA-FLA and PNA-LNA PCR clamp were sensitive enough to detect exon 19 deletions. Two mutations in exon 21 were detected by ASP-PCR only, one L858R substitution was detected only by PNA-LNA PCR clamp. With respect to sensitivity, PCR followed by DNA-FLA achieved a level of detection of at least 10 % of mutated DNA for exon 19 deletion, as for ASP-PCR it was at least 5 % of mutated DNA for L858R substitution. Higher sensitivity of 1 % of mutated DNA was achieved by PNA-LNA PCR clamp technique for both mutations. The use of different methodological techniques authenticates the negative result of molecular tests.