Objective: To develop clinical algorithms that improve the sensitivity of surrogate markers to initiate the use of highly active antiretroviral therapy (HAART) in resource-limited settings.
Design: A retrospective evaluation of total lymphocyte counts (TLC) and hemoglobin to predict the CD4 lymphocyte count.
Methods: A total of 3269 members of the Johns Hopkins HIV observational cohort contributed 22 690 paired observations of CD4 lymphocyte counts and TLC. Two methods were used to evaluate the effect of combining TLC and hemoglobin to predict CD4 cell counts below 200 cells/mm3 before the initiation of HAART in 1451 participants; 55.3% of participants had CD4 cell counts below 200 cells/mm3.
Results: TLC below 1200 cells/mm3 and hemoglobin below 12 g/dl significantly predicted CD4 cell counts below 200 cells/mm3. For TLC alone sensitivity was 70.7% and specificity was 81.7%. For both men and women, we chose a TLC lower cutoff point of 1200 cells/mm3, an upper cutoff point of 2000 cells/mm3, and hemoglobin of 12 g/dl. For men, method I generated sensitivity of 78.0% and specificity of 77.5%. Method II improved specificity to 81.8%. For women, method I increased sensitivity to 85.6% and decreased specificity to 64.1%. Method II improved specificity to 81.4%.
Conclusion: TLC below 1200 cells/mm3 were associated with CD4 cell counts below 200 cells/mm3 as in the WHO guidelines, but sensitivity was low. Adding hemoglobin to TLC increased sensitivity, thereby reducing the risk of false-negative results. Our model may serve as a template for the development of algorithms to initiate the use of HAART in resource-limited settings.