OBJECTIVE This study explores whether viral load measurements can be used in resource-limited settings to target those in need of adherence assistance. It was hypothesized that high plasma viral loads (pVLs) (>/=500 HIV-1 RNA copies/mL) were the result of poor antiretroviral therapy adherence and amenable to improvement with adherence assistance. DESIGN A single-arm, multicentre pilot study was conducted from November 2003 to March 2004 on 606 treatment-experienced patients who had initiated an antiretroviral regimen in Mali and Burkina Faso >/=6 months before study enrolment. In these patients, those whose pVL was >/=500 copies/mL were offered 1 month of modified directly administered antiretroviral treatment (mDAART) with weekly follow-up visits from pharmacists or adherence counsellors. METHODS An adherence questionnaire was given to all cohort patients and viral load was used to screen for patients with >/=500 copies/mL. mDAART participants included cohort patients with >/=500 copies/mL, who completed the adherence questionnaire. Genotypic analyses were conducted on samples taken prior to and after the intervention. The intervention was considered effective when there was a decrease of >/=1 log(10) in pVL. RESULTS mDAART was effective in over one-third of the intervention participants, while in two-thirds no decrease in pVL was observed. The majority of mDAART participants had major resistance mutations. CONCLUSIONS pVL measurement was useful to identify patients who needed adherence assistance. However, because it was performed >/=6 months after starting treatment, mDAART came too late for most participants, as they had already developed important resistance mutations that might have been avoided with better laboratory monitoring.