Application of an algorithm to predict CD4 lymphocyte count below 200 cells/mm(3) in HIV-infected patients in South Africa.

Abstract

We welcome the attempt by Spacek et al. [1] to develop a practical, accurate, and less expensive alternative to universal CD4 lymphocyte testing to guide the initiation of potent antiretroviral therapy in resource-limited settings. In South Africa the government has recently announced its intention to provide antiretroviral treatment for individuals with HIV and AIDS [2]. We are aware that the Operational Plan for Comprehensive HIV and AIDS Care, Management and Treatment for South Africa calls for the initiation of antiretroviral treatment based on clinical assessment and CD4 lymphocyte count. However, universal CD4 lymphocyte testing in South Africa will require substantial upgrading of the current nationwide capacity. Therefore, we investigated the utility of the algorithm of Spacek et al. [1] in our own clinic setting. As flow cytometry is available in South Africa, we chose to apply method II in their paper. Using this algorithm, patients with total lymphocyte counts of less than 1200 cells/mm3 are predicted to have CD4 lymphocyte counts of less than 200 cells/mm3; patients with total lymphocyte counts of 1200 cells/mm3 or greater and hemoglobin levels less than 12 g/dl require CD4 lymphocyte testing; and patients with total lymphocyte counts of 1200 cells/mm3 or greater and hemoglobin levels of 12 g/dl or greater are considered to have CD4 lymphocyte counts of 200 cells/mm3 or greater. We conducted a retrospective review of patients’ charts at the King Edward VIII Hospital Family Clinic, a public sector, tertiary care referral HIV/AIDS Clinic in Durban, KwaZulu, Natal, where potent antiretroviral therapy is still inaccessible to the majority, but where CD4 lymphocyte counts and full blood counts are performed routinely. During the 3-month period ending 30 November 2003, 281 antiretroviral-naive adults (mean age 33 years, 75% women, 93% black), with CD4 lymphocyte counts and full blood counts obtained simultaneously, were evaluated (Table 1). The median CD4 lymphocyte count was 209 cells/mm3 and 48.4% of patients had CD4 lymphocyte counts of less than 200 cells/mm3. Eighty-six patients (30.6%) had total lymphocyte counts of less than 1200 cells/mm3. In this low total lymphocyte count group, 74 patients (26.3%) had CD4 lymphocyte counts of less than 200 cells/mm3 and only 12 (4.3%) had CD4 lymphocyte counts of 200 cells/mm3 or greater. These 12 patients, who would have received antiretroviral therapy prematurely on the basis of the algorithm, had a mean CD4 lymphocyte count of 301 cells/mm3 (range 220–515), and represent false positives. A total of 195 patients (69.4%) had total lymphocyte counts of 1200 cells/mm3 or greater. In this high total lymphocyte count group, 104 patients (37.0%) had hemoglobin levels less than 12 g/dl and would require flow cytometry based on the algorithm. The remaining 91 patients (32.4%) with total lymphocyte counts of 1200 cells/mm3 or greater had hemoglobin levels of 12 g/dl or greater. In this high total lymphocyte count, high hemoglobin group, 71 (25.3%) had CD4 lymphocyte counts of 200 cells/mm3 or greater and only 20 (7.1%) had CD4 lymphocyte counts of less than 200 cells/mm3. These false-negative individuals, with a mean CD4 lymphocyte count of 124 cells/mm3 (range 42–193), may possibly be denied antiretroviral therapy despite having unidentified low CD4 lymphocyte counts. Of note was the fact that a further review of this group's medical records revealed AIDS-defining conditions in nine out of 20 patients, indicating a need for antiretroviral agents irrespective of the CD4 lymphocyte count [3].Table 1: Application of the method II algorithm of Spacek et al. [1] in 281 HIV-infected patients in Durban, South Africa.In conclusion, our findings demonstrate that if the algorithm of Spacek et al. [1] were applied to our setting, 63.0% of staging CD4 lymphocyte tests could be avoided, whereas 7.1% of patients might be inaccurately denied antiretroviral al agents an and 4.3% might be placed on antiretroviral agents prematurely. A similar algorithm developed with local data and incorporating clinical staging might perform even better.