We demonstrated a high overall incidence of any WHO stage 3 or 4 condition (40.02 episodes per 100 person-years) in the 3 years following ART initiation in patients in RLS. In the 3 months immediately following ART initiation, the incidence of any WHO stage 3 or 4 condition was nearly 14 times higher than that reported for a resource-rich setting by the Swiss HIV Cohort Study (106.5 episodes compared with 7.7 episodes per 100 person-years) 
. However, we also showed that there were significant reductions, particularly in the first 12 months, in the incidence of new WHO stage 3 or 4 conditions during the first 36 months of ART. Both stage 3 and 4 conditions reduced in overall incidence by more than 90% in the first year and reductions in disease specific incidence rates of the conditions with highest initial incidence ranged from 82% to 97%. These striking decreases in incidence during the first year of ART are similar to those observed in patients on ART in high income countries 
. As our analyses are not adjusted for death and loss to follow-up in the cohort, the absolute reductions in incidence rates cannot be ascribed to the beneficial effects of ART alone 
. Nevertheless it is known that the immune restorative effects of ART reduce the incidence of new or recurrent WHO stage 3 and 4 conditions 
and it is likely that ART has played a significant role in their reductions in our cohort. This emphasizes the importance of access to early ART initiation for those eligible to reduce the impact of HIV related conditions during early ART in RLS. This requires increased efforts to enable early diagnosis and accurate clinical and immunological staging of HIV infected patients, and their rapid access to effective, non-toxic, free and sustained ART in these settings.
The considerable incidence of HIV associated clinical conditions during early ART also emphasizes the importance of increased disease prevention efforts prior to ART commencement. Such interventions include isoniazid chemophrophylaxis for TB 
, fluconazole prophylaxis for those who are cryptococcal antigen negative and have delays in ART initiation 
and cotrimoxazole therapy for PCP, toxoplasmosis, bacterial sepsis and malaria 
. Resources and improved tools are also required to facilitate early and improved diagnosis of opportunistic infections, as well as effective, adapted and affordable treatments to minimize their impact on morbidity and mortality.
We defined the burden of different WHO stage 3 and 4 conditions stratified by age and region. This provides information for clinicians to use in determining which conditions to focus on for prevention, early diagnosis and treatment. It is also vital information for programme planners to guide allocation of resources required for disease management in patients on ART, and for strategic planning and resource prioritization. For example, age-specific and regional incidence rates of TB in patients on ART can aid strategic decision making about which patient populations to prioritise in implementing isoniazid preventive therapy (IPT) if universal IPT is not chosen or possible. Also, knowing specific disease incidences can aid in decision making about programme drug formularies.
TB and candidiasis were the most common conditions seen in our RLS cohort. TB is a leading cause of disease in HIV infected patients and is the leading cause of HIV-related deaths 
. Our results re-affirm the importance of TB as a pathogen in HIV infected patients treated with ART. Importantly, our results support other studies in suggesting that ART alone reduces the incidence of TB 
. Provision of IPT can further reduce the impact of TB whilst on ART 
and is currently recommended 
. Likewise our study showed ART is associated with a reduction in the incidence of candidiasis. Although a recent study from Uganda showed that giving fluconazole prophylaxis to those with a CD4 count <200 cells/mm3
can further reduce the impact of candidiasis both pre and post ART 
, it did not reduce overall mortality, and is currently not recommended by WHO 
. Important contributions to the disease burden in our study were also made by potentially preventable stage 4 conditions such as non-tuberculous mycobacteria, cryptococcosis, pneumocystis and toxoplasmosis. The most common conditions differed from those in a resource rich setting where the oseophageal candidiasis, PCP and cytomegalovirus disease were the most frequently reported conditions during the first 6 months of ART 
There were important regional differences in the overall and disease specific incidences of any stage 3 and 4 condition following ART initiation. The incidence of any stage 3 HIV associated condition was almost 4 times higher in Africa than Asia despite African patients being less immunosuppressed overall at baseline. This may be due to increased exposure to the relevant pathogens in Africa or less likely, to improved diagnosis in Africa 
. Some individual stage 4 conditions were more common in Asian programmes, particularly in the first 3 months after ART initiation, likely due to the higher rate of immunosuppression of these patients at baseline, and differential prevalence of specific pathogens. For example, penicilliosis is not endemic in Africa.
The incidence of any first stage 3 condition did not differ by age however there were important differences in the incidence of individual conditions by age. Pulmonary TB was diagnosed less commonly in children <5 years, although this may be underestimated given the difficulty of diagnosis in this age-group. Additionally, severe bacterial pneumonia was more common in children <15 years compared with adults.
There were significant age-related differences in the overall incidences of stage 4 conditions that may be explained by the relative level of baseline immunodeficiency. HIV-infected children <5 years usually have rapidly progressive disease with high mortality 
. This increased immunodeficiency may explain the higher incidence of stage 4 conditions in this age-group. Conversely, many children in the 5–14 age-group will have been infected peripartum, but survived due to less rapidly progressive disease. This may explain this group’s low incidence of stage 4 conditions. In addition, there were more differences seen in the incidence of specific stage 4 conditions between children <5 years and adults. Children <5 years were more likely to have HIV wasting compared with adults but less likely to have all other conditions except pneumocystis pneumonia. There were no cases of toxoplasmosis, penicilliosis or Kaposi’s sarcoma in patients <5 years of age. As reported previously 
, crytptococcosis was rare under 15 years of age (1 case), supporting recent WHO guidelines recommending against screening and pre-emptive treatment in children due to low prevalence 
Information about age-specific incidence rates of various conditions during ART may have important practical applications. For example, TB incidence reduced by 48% during the first 12 months of ART in children <5 years which was less than the 79% reduction in adults and 77% reduction in children 5–14 years. This suggests IPT may have relatively greater additional impact in children <5 years than older age-groups during early ART. In addition, the incidence of bacterial pneumonia during the first 12 months of ART was up to 7 times higher in children than adults, supporting particular attention being paid to the prevention, diagnosis and treatment of this condition in children.
This study has a number of limitations. Firstly, reliance on basic investigations to support clinical assessment in the diagnosis of HIV associated conditions may have resulted in diagnostic inaccuracies. For example TB in children <5 years old is difficult to confirm with AFB stains alone, and reliance on clinical assessment may have under or overestimated its true incidence. In addition, clinician practices and availability of investigations may have varied between programmes, influencing the relative frequencies of diagnoses. Nevertheless, the results of this study remain relevant because the diagnoses reflect the reality of clinical practice in most RLS where ART is delivered and involves a large cohort of patients. Secondly, the lack of pre-ART data precluded the comparison of incidence rates before and after ART. Thirdly, the Asian data was predominantly from Myanmar programmes so results may not be generalisable across all Asian programmes. Fourthly, we cannot judge the extent to which declining incidence rates are due to ART and care effects; some of these declines may be due to deaths of severely ill persons in the prior time periods. Finally, we were unable to ascertain whether patients lost to follow-up developed any HIV related conditions. Furthermore, as inclusion criteria required at least 3 months of follow-up on ART, patients developing stage 3 and 4 conditions in the first 3 months, which may have increased their risk of lost-to follow-up, could have led to an underestimation of incidence rates in this initial period.
In RLS the incidence of WHO stage 3 and 4 conditions in the early period following ART initiation is high when compared to similar studies in resource-rich settings. However, with early deaths of the sickest patients and with the benefits of ART for the remaining patients, the incidence of HIV associated conditions greatly reduces with time on ART, particularly within the first 12 months. TB and candidiasis account for the greatest disease burden, but many other potentially preventable diseases are also important. Efforts directed towards prevention, early diagnosis and effective treatment of these conditions before and during ART, and the early and accurate assessment for, and initiation of, ART in eligible patients, are important to reduce morbidity during ART in RLS.