Repeated measurements on persons infected with HIV‐1 indicate that infection has a dynamic impact on several markers of immune suppression and activation. The objectives of this report are: )a( to provide a statistical model for the correlation structure of serial measurements of immunological markers, and )b( to identify features of marker profiles associated with the timing of AIDS diagnoses. We analyse data obtained from 328 seroconverters participating in the Multicenter AIDS Cohort Study on whom the date of HIV‐1 seroconversion is known within ∓ 4·5 months. Immunological markers considered here are CD4 cell counts, serum β2‐microglobulin and serum neopterin. The statistical model for HIV‐related changes in markers consists of )1( a piecewise linear regression model for the trajectories of markers over time and )2( a two‐parameter autocorrelation function that generalizes Markovian and simple random effects autocorrelation structures. Application of this model for marker measurements revealed a high degree of tracking, as the estimated autocorrelation function exhibited sub‐exponential decay over time. Though current marker levels are most informative on future values, there is substantial information )memory( in previous measurements. A feature suggested by the analysis of groups formed according to the length of the AIDS‐free period, is the sequential divergence of the CD4 trajectories where steeper declines occurred with a two‐year lag prior to AIDS onset. For AIDS cases diagnosed 3–5 and 5–7 years after seroconversion, the rates of decline compared with those free of AIDS for at least 4 years were steeper by 95 and 46 per cent respectively at two years prior to AIDS.
ASJC Scopus subject areas
- Statistics and Probability