CD4 T-lymphocyte percentages corresponding to CD4 T-lymphocyte count thresholds in a new staging system for HIV infection.

To the Editors: For epidemiologic surveillance of HIV infection in the United States, until this year, the staging system for adults (published in 2008) had been separate from the classification system for children (published in 1994).1,2 To design a single staging system for both adults and children based primarily on absolute CD4 T-lymphocyte counts, we retained the age-specific CD4 count thresholds used to define the boundaries between stages 1, 2, and 3 (called “immunologic categories” rather than “stages” in the 1994 classification for children). Values greater than or equal to the upper threshold indicate stage 1, values less than the upper threshold but greater than or equal to the lower threshold indicate stage 2, and values less than the lower threshold indicate stage 3 (AIDS). For children aged <1 year, the lower and upper CD4 count thresholds are 750 and 1500 (cells/μL); for children aged 1 to <6 years, they are 500 and 1000; for children aged 6 to <13 and for adults and adolescents aged 13 or older, they are 200 and 500. Those staging/classification systems used both the absolute CD4 count and the CD4 percentage of total lymphocytes to classify cases into stages; if the CD4 count and the CD4 percentage indicated different stages, the more advanced of the 2 stages was selected. If one of these measurements was not available, the classification was based solely on the other measurement. The lower and upper CD4 percentage thresholds in those staging/classification systems were 15% and 25% for all 3 age groups of children, and 14% and 29% for adults and adolescents.1,2 In developing an updated staging system, we reassessed the relationship between the CD4 counts and the CD4 percentages and selected the mean CD4 percentage corresponding to each CD4 count threshold. We analyzed pairs of CD4 counts and CD4 percentages from 2 data sources: (1) the National HIV Surveillance System (NHSS) of the Centers for Disease Control and Prevention (data received by March 31, 2013) and (2) the HIV Research Network (HIVRN), a consortium of 19 facilities providing care to HIV-infected patients. Because data from these 2 sources may overlap to an unknown extent, we analyzed them separately to avoid duplicating observations in a single analysis. After testing several models of the relationship between CD4 counts and CD4 percentages, we concluded that a linear regression model of their natural logarithmic transformations would be best, with the logarithm of the CD4 percentage as the dependent variable predicted by the logarithm of the CD4 count: log (CD4 percentage) = Intercept + Slope × log (CD4 count). We calculated the mean CD4 percentage predicted by each CD4 count threshold by placing each CD4 count threshold in the regression equation with the estimated Intercept and Slope (regression coefficient), and taking the antilogarithm of the result. The results from HIVRN data were similar to those from NHSS data (Table 1). Rounded to the nearest whole percentage, the lower and upper CD4 percentage thresholds derived from NHSS data were as follows: For children aged <1 year: 26% and 34%. For children aged 1 to <6 years: 22% and 30%. For children aged 6 to <13 years: 14% and 24%. For adults and adolescents aged ≥13 years: 14% and 26%. TABLE 1: Mean CD4 T-Lymphocyte Percentages Corresponding to Absolute CD4 T-Lymphocyte Count Thresholds for Staging, by Age Group, as Predicted by Regression Model*The corresponding results derived from HIVRN data were as follows: For children aged <1 year: 26% and 32%. For children aged 1 to <6 years: 22% and 29%. For children aged 6 to <13 years: 13% and 23%. For adults and adolescents aged ≥13 years: 14% and 25%. Thus, the upper CD4 percentage threshold of 29% for adults and adolescents, and the lower and upper thresholds of 15% and 25% for children aged <6 years in the systems that had been used through 2013 no longer seem to correspond accurately to the CD4 count thresholds for which they had been substituted. This is probably due to the number of observations in our analysis being larger than the number used in the earlier analyses on which the old thresholds were based. The CD4 percentages corresponding to the CD4 count thresholds could alternatively be derived by using a model in which the logarithm of the CD4 count is the dependent variable and the logarithm of the CD4 percentage is the independent variable, instead of the reverse. Then, the model equation could be solved for log (CD4 percentage) = [log (CD4 count) − Intercept]/Slope, and the values for the CD4 count threshold, intercept, and coefficient could be plugged in, and the antilogarithm would yield the corresponding CD4 percentage. This alternative method yielded results slightly different from those presented here. For simplicity, we chose not to use those alternative results for selecting the CD4 percentage thresholds in our staging system. In the updated staging system, we will base the CD4 percentage thresholds on the results from the models using the NHSS data rather than the HIVRN data because the former are more nationally representative than the latter. Despite the small but statistically significant differences between the upper CD4 percentage thresholds derived for children aged 6 to <13 years (24%) and adults/adolescents aged ≥13 years (26%), we plan to combine these 2 age groups in the staging system and apply the latter CD4 percentage threshold (26%) to both. This simplification is likely to misclassify only a small proportion of children aged 6 to <13 years with a CD4 percentage ≥24% but <26% into stage 2 instead of stage 1, because the upper CD4 count thresholds for these 2 age groups are identical (500 cells/μL for both) and CD4 percentages will be used for staging only if the corresponding CD4 counts are unavailable. Of the 5,671,947 CD4 test results reported to NHSS by March 31, 2013, 8% were CD4 percentages unaccompanied by CD4 counts, and 4% were CD4 counts unaccompanied by CD4 percentages. We will give priority to the CD4 count over the CD4 percentage because clinical evidence suggests that the CD4 percentage has little effect on prognosis after adjusting for the CD4 count.3,4 The new rule giving priority to the CD4 count over the CD4 percentage will avoid the exaggeration of the number of cases with advanced disease that was made by the previous rule that based the stage on whichever result (CD4 count or CD4 percentage) indicated the more severe stage. Compared with the numbers that would result from giving priority to the CD4 count, the previous rule under-counted stage-1 cases by 16% and over-counted stage-3 cases by 6% overall. The updated staging system will use additional criteria, such as a negative HIV test result within 180 days before a positive HIV test result as an indicator of stage 0 (early infection), and the diagnosis of an opportunistic illness as an alternative indicator of stage 3, but these details are beyond the scope of the issues related to CD4 test results discussed here. The updated staging criteria were implemented when a revised surveillance case definition for HIV infection was published in 2014.5 The staging system described here is intended primarily for public health surveillance of HIV infection on the population level. Health departments can use the staging system to evaluate prevention and care by analyzing the trends in the distributions of reported cases by stage at diagnosis and the speed of progression to more advanced stages after diagnosis. This staging system is not intended to guide the clinical management of individual patients. US panels on antiretroviral guidelines recommend initiating antiretroviral therapy for all HIV-infected adults, adolescents, and children regardless of CD4 counts or CD4 percentages.6–8 ACKNOWLEDGMENTS The authors thank Ye Cui, PhD, for help with SAS programming for the statistical analysis, and Orlando J. Davy for exploratory data analyses.