Setting: The network of supranational tuberculosis reference laboratories (SRLs).
Objective: To evaluate the annual SRL Rounds 6-14 of proficiency testing for first-line drug susceptibility testing (DST).
Design: Panels consisted of 20-30 cultures (including 10 pairs of duplicate strains), aiming at 50% resistance prevalence with a variety of profiles. The 27 SRLs participating in at least one of these rounds were free to use their preferred DST method. A judicial gold standard of at least 80% concordant 'susceptible' or 'resistant' was used to determine sensitivity, specificity and efficiency; otherwise the strain was excluded.
Results: Of 600 strains, 10% were excluded from evaluation. The average SRL sensitivity and specificity varied between rounds, without attaining significance or trends. Both sensitivity and specificity remained at >95% for isoniazid (INH), rifampicin (RMP) and streptomycin and at >80% for ethambutol. The 16 SRLs participating in all rounds performed consistently better.
Conclusion: The rounds succeeded in comparing the proficiency of laboratories, and should be further promoted for DST quality assessment. However, to function with greater precision and to ultimately improve the clinical relevance of DST, the INH and RMP judicial result gold standard also needs to take into account genotypic and treatment outcome information.