Background: A sensitive, phenotypic reverse transcriptase (RT)-based drug susceptibility assay for the detection of etravirine (ETR) resistance in patient isolates was developed and compared with the results from direct sequencing and ultra-deep pyrosequencing (UDPS). Methods: Samples were obtained from 15 patients with antiretroviral therapy (ART) failure and from five non-nucleoside reverse transcriptase inhibitor (NNRTI)-naïve patients of whom four were infected by an NNRTI-resistant strain (transmitted drug resistance, TDR). In five patients, two consecutive samples (a and b) were taken for follow up of the virological response. HIV-1 RT was purified and drug susceptibility (IC 50) to ETR was estimated. Direct sequencing was performed in all samples and UDPS in samples from nine patients. Results: Increased IC 50 to ETR was found in samples from 13 patients where direct sequencing predicted resistance in only four. UDPS identified additional (N = 11) NNRTI resistance associated mutations (RAMs) in six of nine tested patients. During early failure, IC50 increases were observed in three of six patients without any ETR-RAMs detected by direct sequencing. In further two patients, who stopped NNRTI before sampling, increased IC50 values were found shortly after, despite absence of ETR-RAMs. In two patients who had stopped NNRTI for >1 year, a concordance between phenotype and genotypes was found. Two patients with TDR had increased IC50 despite no ETR-RAMs were detected by direct sequencing. UDPS revealed additional ETR-RAMs in four patients with a discrepancy between phenotype and direct sequencing. Conclusions: The RT-based phenotypic assay showed decreased ETR susceptibility in patients where direct sequencing predicted ETR-sensitive virus. This increased phenotypic sensitivity was to a large extent supported by UDPS and treatment history. Our method could be valuable for further studies on the phenotypic kinetics of NNRTI resistance. The clinical relevance remains to be studied in larger patient-populations. © 2014 Agneskog et al.