Impaired linear growth and slower pubertal growth can be associated with perinatal HIV infection. We characterised growth relative to population norms, among the full adolescent period in southern Africa to better understand processes leading to morbidity in adulthood. We conducted a secondary analysis of 945 adolescents aged 8-20 years from urban Malawi and Zimbabwe; we included children with HIV (CWH), an uninfected comparison group from a cohort study, and CWH with co-morbid chronic lung disease (CLD) from a randomised controlled trial. We used latent class analysis of anthropometric Z-scores generated from British 1990 reference equations at two annual time-points, to identify growth trajectory profiles and used multinomial logistic regression to identify factors associated with growth profiles. Growth faltering (one or more of weight-for-age, height-for-age, or BMI-for-age Z-scores < -2) occurred in 38% (116/303) of CWH from the cohort study, 62% (209/336) of CWH with CLD, and 14% (44/306) of HIV-uninfected participants. We identified seven different growth profiles, defined, relatively, as (1) average growth, (2) tall not thin, (3) short not thin, (4) stunted not thin, (5) thin not stunted, (6) thin and stunted and (7) very thin and stunted. Females in profile 3 exhibited the highest body fat percentage, which increased over 1 year. Males at older age and CWH especially those with CLD were more likely to fall into growth profiles 4-7. Improvements in height-for-age Z-scores were observed in profiles 6-7 over 1 year. Interventions to target those with the worst growth faltering and longer-term follow-up to assess the impact on adult health are warranted.