In the 1970s, survival rates after treatment for acute lymphoblastic leukaemia (ALL) in children and young adults (less than 25 years) in India were poor, even in specialised cancer centres. The introduction of a standard treatment protocol (MCP841) and improvements in supportive care in three major cancer centres in India led to an increase in the event-free survival rate (EFS) from less than 20% to 45-60% at 4 years. Results of treatment with protocol MCP841 between 1984 and 1990 have been published and are briefly reviewed here. In addition, previously unpublished data from 1048 patients treated between 1990 and 1997 are reported. Significant differences in both patient populations and treatment outcome were noted among the centres. In one centre, a sufficiently large number of patients were treated each year to perform an analysis of patient characteristics and outcome over time. Although steady improvement in outcome was observed, differences in the patient populations in the time periods examined were also noted. Remarkably, prognostic factors common to all three centres could not be defined. Total white blood cell count (WBC) was the only statistically significant risk factor identified in multivariate analyses in two of the centres. Age is strongly associated with outcome in Western series, but was not a risk factor for EFS in any of the centres. Comparison of patient characteristics with published series from Western nations indicated that patients from all three Indian centres had more extensive disease at presentation, as measured by WBC, lymphadenopathy and organomegaly. The proportions of ALLs with precursor T-cell immunophenotypes, particularly in Chennai, were also increased, even when differences in the age distribution were taken into consideration (in <18-year olds, the range was 21.1-42.7%), and in molecular analyses performed on leukaemic cells from over 250 patients less than 21-years-old with precursor B-cell ALL, a lower frequency of TEL-AML1-positive ALL cases than reported in Western series was observed. The worse outcome of treatment in Indian patients compared with recent Western series was probably due to the higher rate of toxic deaths in the Indian patients, and possibly also due to their more extensive disease - which is, at least partly, a consequence of delay in diagnosis. Differences in the spectrum of molecular subtypes may also have played a role. The higher toxic death rates observed are likely to have arisen from a combination of more extensive disease at diagnosis, co-morbidities (e.g., intercurrent infections), differences in the level of hygiene achievable in the average home, poor access to acute care, and more limited supportive care facilities in Indian hospitals. Toxic death was not associated with WBC at presentation, and hence would tend to obscure the importance of this, and, potentially, other risk factors, as prognostic indicators. Since the prevalence of individual risk factors varies in different populations and over time, their relative importance would also be expected to vary in different centres and in different time periods. This was, in fact, observed. These findings have important implications for the treatment of ALL in countries of low socioeconomic status; it cannot be assumed that risk factors defined in Western populations are equally appropriate for patient assignment to risk-adapted therapy groups in less affluent countries. They also demonstrate that heterogeneity in patient populations and resources can result in significant differences in outcome, even when the same treatment protocol is used. This is often overlooked when comparing published patient series.