Background: Data on changes over time in HIV tropism and association with disease progression in perinatal HIV-1 infection (PaHIV) is sparse. Optimal sequencing of antiretrovirals (ART) including CCR5 antagonists is a priority for a young population currently requiring life long ART and requires a clear understanding of the natural history of coreceptor usage in PaHIV. Methodology: Eligible patients with PaHIV were CCR5 antagonist naive and were defined as (1) slow (ART from 10+ yrs) or (2) rapid (ART pre 5 yrs) progressors and (3) long term non-progressors (LTNP). RNA was extracted from stored plasma samples at baseline and at yearly time points or virological failure to latest follow up. The V3 region of gp120 was sequenced and tropism determined using the Geno2pheno algorithm (FPR 5.75%). Logistic regression with generalised estimating equations was used to assess factors associated with the presence of R5 virus. Time to tropism change was assessed using standard survival methods. Results: At baseline (n=48) median age was 12 yrs, 52% female, 79% black African, 96% non-B subtype, group 1(n=20), 2(n=17) and 3(n=11). 81% (39/48) had R5 and 19% (9) had X4/dual-mixed (DM) viruses. Median follow-up was 7.7 yrs (308.6 person-yrs) with a median of 5 (range 1-14) samples per subject (252 samples total). Analysing all samples, the presence of R5 virus was associated with higher current CD4 count (median 520 cells/μL for R5 virus vs 202 for X4; p=0.0005), LTNP (35% vs 11% p=0.05), non Black African ethnicity (74% vs. 89%; p=0.05) and female (55% vs 28%; p=0.005). Of 38 patients with R5 virus at baseline and at least one follow-up sample, switch to X4/DM occurred in 12 (30.8%). Estimated 5 yr risk of tropism switch to X4/DM was 24.4% (95% CI 9.7-39.2%). Lower current CD4 predicted coreceptor switch (unadjusted HR=0.62 per 50 cells higher; 95% CI 0.47-0.81; p=0.0006) but not viral load (p=0.81), age (p=0.85) or clinical group (p=0.56). Of 19 patients who ever had X4/DM virus and at least one further sample, 11 (58%) had R5 virus at one or more subsequent time points. 5/11 had three or more switches between R5 and X4/DM over time, 3 of whom were R5 at last follow up. Conclusions: As seen in adult studies, lower CD4 count was predictive of coreceptor switching in PaHIV. CCR5 antagonists were a treatment option for 81% at 12 yrs, falling to 56% over 7.7 yrs of follow up. Globally 30% of children with PaHIV are triple class experienced and currently are expected to remain on ART for life. Paediatric studies of CCR5 antagonists should be expedited to ensure they are a treatment option before tropism switching occurs in children. The frequent reversion from X4 to R5 may result in CCR5 antagonists being prescribed to patients harbouring undetected X4 virus and careful monitoring of individuals commencing such therapy is indicated.