This retrospective analysis of routine programme data from Mbagathi District Hospital, Nairobi, Kenya shows the difference in rates of loss to follow-up between a cohort that paid 500 shillings/month (approximately US$7) for antiretroviral drugs (ART) and one that received medication free of charge. A total of 435 individuals (mean age 31.5 years, 65% female) was followed-up for 146 person-years: 265 were in the 'payment' cohort and 170 in the 'free' cohort. The incidence rate for loss to follow-up per 100 person-years was 47.2 and 20.5, respectively (adjusted hazard ratio 2.27, 95% CI 1.21-4.24, P=0.01). Overall risk reduction attributed to offering ART free of charge was 56.6% (95% CI 20.0-76.5). Five patients diluted their ART regimen to one tablet (instead of two tablets) twice daily in order to reduce the monthly cost of medication by half. All these patients were from the payment cohort. Payment for ART is associated with a significantly higher rate of loss to follow-up, as some patients might be unable to sustain payment over time. In resource-limited settings, ART should be offered free of charge in order to promote treatment compliance and prevent the emergence of drug resistance.