Several guidelines are available to guide the initiation of highly active antiretroviral therapy (HAART) in human immunodeficiency virus (HIV)-infected children. The recommendations in these guidelines show significant variability. Because there is no well-established evidence on when to start HAART, it is left to the discretion of the pediatrician which guidelines to follow. We conducted a survey concerning the indications for starting antiretroviral therapy among pediatricians involved in the treatment of HIV-infected patients in Europe and the United States. We compared the results of this survey with the guidelines available at the time, the recently adapted guidelines and literature evidence. Our results indicate that in clinical practice HAART was initiated at higher viral loads and lower CD4 counts than recommended by the guidelines. American guidelines recommended and still recommend more aggressive treatment than the European guidelines, and this is reflected in clinical practice. Until recently all guidelines were based on long term risk analyses of progression to acquired immunodeficiency syndrome (AIDS) and death performed in cohort data. A recent short term risk analysis makes it possible to calculate the 6 or 12-month risk for progression to AIDS or death for an individual child. Because viral load and CD4 count are typically measured every 3 months, one can argue that it is clinically more relevant to base the decision of when to start HAART on the short term probability of disease progression. Guidelines in Europe are now based on this type of analysis. The American guidelines only adopted the thresholds for CD4 and viral load. The short term risk analysis also shows that the risk for developing AIDS varies markedly with age. This should be reflected in all guidelines. Determining the acceptable risk of disease progression is difficult and influenced by patient-, doctor- and culture-related factors. The controversy over whether or not to treat asymptomatic infants is unresolved as well. All infants have a very high risk of disease progression regardless of their viral load or CD4 count, but lifelong treatment with a potential for significant toxicities and risk of developing resistance is also not an appealing option. We recommend an attempt to achieve a consensus among the different working groups to reduce the number of different guidelines, which should be based on the literature evidence. Because all risk analyses are based on information from the pre-HAART era, a head-to-head trial comparing early versus deferred HAART would be useful. This may be difficult to accomplish. The first step could be an analysis of retrospective data from collaborative cohort data.