Abstract Body

Background: CD4 cell count is a key measure of HIV disease progression, and has been the basis of successive international guidelines for treatment initiation. CD4 cell dynamics are used in mathematical and econometric models for evaluating public health needs and interventions.

Methods: Here, we analyse a subset of the ATHENA cohort, including individuals with date of infection estimated to within one year and with extensive and intensive clinical follow up. Because CD4 counts are intrinsically noisy, we separate the analysis of individual CD4 dynamics over time into a model of long-term trends of smoothed CD4 counts, and an observation model that relates actual CD4 measurements to the underlying smoothed counts. We use a monotonic spline smoothing model, and based on smoothed counts derive the average time to CD4 thresholds CD4≤500, CD4≤350, CD4≤200 cells/mm3 as well as the average time to death, and the proportion of individuals starting in each category after seroconversion. We examine individual-level co-factors which influence these rates. We perform the analysis in both newly infected individuals that have not received antiretroviral therapy (ART) and those who have interrupted therapy for at least six months.

Results: Amongst untreated individuals, the time spent in each compartment was on average 3.09 (CD4>500 cells/mm3), 2.49 (CD4 350-500), 4.95 (CD4 200-350) and 1.07 (CD4≤200) years. Only 77% of individuals had CD4 >500 cells/mm3 at or shortly after seroconversion. Set-point viral load (SPVL) was an important determinant of CD4 progression; individuals with ≥5 log10 copies/ml took 5.32 years to reach CD4≤200 cells/mm3 compared to 11.53 years for SPVL <4 log10 copies/ml (see table). SPVL was not an important predictor of progression after treatment interruption, and CD4 dynamics were otherwise similar to pre-ART dynamics, providing evidence of true immune reconstitution during ART.

Conclusions: Our analyses show that many individuals already have CD4≤500 cells/mm3 at or shortly after seroconversion, and that set point viral load strongly influences initial CD4 cell count as well as rate of CD4 decline. Hence guidelines on treatment initiation should consider criteria based on both current CD4 count and viral load., while mathematical models should incorporate SPVL stratification. Our study also provides new estimates of CD4 dynamics after interruption of ART, which could be used in such models.

⇥

⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥

⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥

⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥

⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥

⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥⇥

⇥⇥

⇥

  From infection From CD4 >500 From CD4 350-500 From CD4 200-350
logSPVL < 4.0 11.53 (9.39-13.18) 12.26 (10.06-13.90) 7.46 (5.47-9.10) 4.49 (2.72-6.07)
logSPVL 4.0-4.5 10.19 (7.60-11.63) 10.95 (8.34-12.47) 7.89 (5.35-9.36) 5.23 (2.84-6.64)
logSPVL 4.5-5.0 7.94 (6.19- 9.07) 8.69 (6.88- 9.92) 6.44 (4.72-7.59) 4.23 (2.56-5.38)
logSPVL > 5.0 5.32 (3.96- 6.34) 5.80 (4.36- 6.82) 4.44 (3.08-5.55) 2.98 (1.66-4.04)