Adenoid Enlargement and CD4 Counts in HIV-Positive Patients

We read with interest the article by Yousem et al in which the size of adenoids in 21 HIV-positive patients was studied as a function of other clinical parameters (1). We are interested in this topic because our experience indicates that large adenoids usually are seen in patients with relatively high CD4 counts. We, therefore, were surprised to learn that Yousem et al's study reported no significant correlation between the number of CD4 antigens and adenoid size. This was particularly surprising because their results were in apparent conflict with the findings reported by Zadanski et al of 33 HIV-positive patients (2). Because of the differing conclusions, we studied 107 HIV-positive patients at our institution to summarize our experience, and to attempt to understand what factors may underlie the apparently disparate results of these other two studies.

Adenoid diameters were measured on sagittal T1-weighted brain MR images, and correlated with CD4 counts in the same manner described by Yousem et al (1). Correlations were recalculated after clustering the data by adenoid size (< 5 mm, 5–7 mm, 8–12 mm, and > 12 mm), similar to the approach used by Zadanski et al. When analyzed, using Yousem's approach, our data similarly indicated no significant correlation between adenoid size and CD4 counts. When analyzed, using Zagdanski's technique, the data suggested a positive correlation between size and CD4 counts (r = 0.84), supporting the conclusions of those authors. Thus, our data support both opposing conclusions, depending on the technique used to analyze the data. Nonetheless, important deficiencies exist in both of these techniques, which when resolved, permit a reasonable understanding of the relationship between CD4 counts and adenoid size.

Using the method employed by Zagdanski, all data are compressed into only four data points. By compressing the 107 observations into four “pseudo observations,” the inherent variability of the data set is lost, limiting the ability to assess errors or trends. Also, the selection of four groups is arbitrary and allows for artificially easy linear regression line fitting. After all, it would be impossible not to find a high correlation had only two groups been selected. Finally, there is no weighting to account for the large variation in the number of patients assigned to each group.

Although Yousem's approach importantly retains the intersubject variation inhererent to the data set, inspection of the resulting regression plot indicates that, for many adenoid sizes, CD4 counts do not distribute normally about the regression line (Fig 1A). Because testing associated with standard linear regression analysis assumes a normal distribution of data, an F test (and significance testing) is inappropriate for the data set in question until a normalizing transformation is made.

• Download figure
• Open in new tab
• Download powerpoint

To normalize the data distribution, the log of CD4 counts was plotted against adenoid size. Using this transformation, the data points distribute symmetrically about the regression lines (Fig 1B), and modest correlation is obtained that is highly significant (r = .38; P < .0001). Thus, the data support the notion that adenoid size correlates with CD4 counts.

Our analysis of a relatively large group of patients also helps resolve the apparent conflict between the previous studies. The study supports Yousem et al's conclusion that there is not a simple linear relationship between adenoid size and CD4 counts in HIV disease, but we have refined the viewpoint. At the same time, our data support the view of Zagdanski et al that intense adenoid “hypertrophy” tends to occur in patients with relatively high CD4 counts.