Abstract:
BACKGROUND: Antiretroviral drug resistance is becoming increasingly common with the expansion of human
immunodeficiency virus (HIV) treatment programmes in high prevalence settings. Genotypic resistance testing
could have benefit in guiding individual-level treatment decisions but successful models for delivering resistance
testing in low- and middle-income countries have not been reported.
METHODS: An HIV Treatment Failure Clinic model was implemented within a large primary health care HIV
treatment programme in northern KwaZulu-Natal, South Africa. Genotypic resistance testing was offered to adults
(≥16 years) with virological failure on first-line antiretroviral therapy (one viral load >1000 copies/ml after at least 12
months on a standard first-line regimen). A genotypic resistance test report was generated with treatment
recommendations from a specialist HIV clinician and sent to medical officers at the clinics who were responsible for
patient management. A quantitative process evaluation was conducted to determine how the model was
implemented and to provide feedback regarding barriers and challenges to delivery.
RESULTS: A total of 508 specimens were submitted for genotyping between 8 April 2011 and 31 January 2013; in
438 cases (86.2%) a complete genotype report with recommendations from the specialist clinician was sent to the
medical officer. The median turnaround time from specimen collection to receipt of final report was 18 days
(interquartile range (IQR) 13–29). In 114 (26.0%) cases the recommended treatment differed from what would be
given in the absence of drug resistance testing. In the majority of cases (n = 315, 71.9%), the subsequent treatment
prescribed was in line with the recommendations of the report.
CONCLUSIONS: Genotypic resistance testing was successfully implemented in this large primary health care HIV
programme and the system functioned well enough for the results to influence clinical management decisions in
real time. Further research will explore the impact and cost-effectiveness of different implementation models in
different settings.