Background: Clinical trials have led to important changes in the treatment of paediatric malignancies. As a result the current 5-year-survival rate of children with a malignancy is approximately 70%. To date, more than 600 randomised controlled trials (RCTs) have been performed in paediatric oncology. In the process of preparing systematic reviews in this field it is crucial to gain insight in possible biases in these RCTs.
Objective: To determine the methodological quality of RCTs in paediatric oncology published in the years 2001 and 2002.
Methods: MEDLINE was searched for RCTs in paediatric oncology using appropriate MeSH terms, text words and sensitive RCT filters. To determine the methodological quality, for each included study data with regard to concealment of allocation, blinding of patients, care providers and outcome assessors, the use of an intention-to-treat analysis and completeness of follow-up were collected and assessed by 3 independent investigators. Discrepancies between investigators were resolved by discussion. Each adequately applied quality item was graded with 1 point, the quality score could therefore range from 0 to 5.
Results: Sixty-three articles qualified for inclusion; they were published in 24 different journals. The mean number of randomised patients was 192 (range 10 -1204). A majority of the RCTs (54%) described patients with a haematological malignancy, mainly leukaemia. In 50% of the RCTs anti-tumour treatment was the intervention under investigation. The most frequently used outcome measure was (long-term) survival (33%). Adequate concealment of allocation was described in 15 articles (23.8%). Blinding of patients and care providers was applied in 1 article (1.6%). Blinding of outcome assessors was reported in 3 (4.8%) articles. Adequate intention-to-treat analysis took place in 20 (32%) of the trials. Adequate follow-up with an outcome assessment of more than 80% of randomised patients was achieved in 53 articles (84.1%). The mean score of the quality items was 1.46 (range 0-3); none of the articles adequately reported all quality items. In a multivariate regression analysis no significant relation was found between the methodological quality and the number of randomised patients, the impact factor of the journal, publication in a paediatric vs. other journal, whether the journal had explicitly adopted the CONSORT statement or whether the results of the study showed statistical significant differences.
Conclusions: The methodological quality of RCTs in paediatric oncology published in the years 2001 and 2002 seems to be poor. This can be due to poor reporting, failures in the design of the studies, or both. Efforts should be aimed at improving both the quality of reporting and the methodological quality of the design and execution of RCTs in this field.