Financial relationships between industry, researchers and academic institutions are becoming increasingly complex, raising concern about sponsors' involvement in the conduct of biomedical research. A review of published randomised trials (RCTs) in cancer research was performed to assess adherence to the 1997 disclosure requirements and to document the nature of the disclosed interests. Source(s) of study support, author-sponsor relationships and the role of the study sponsor were assessed for all RCTs published between 1999 and 2003 in 12 international journals. A total of 655 cancer RCTs were identified. Of these, 516 (78.8%) disclosed the source of sponsorship. The nature of the relationship between the authors and the study sponsor was included in 219 of the 227 industry-sponsored studies. The most commonly cited relationships were (131 studies had multiple relations): grants (93.6%); employment (39.2%); consultant/honorarium (12.7%) and stock ownership and participation in a speaker's bureau (12, 5.5% each). Only 41 (18%) of the 227 industry-sponsored RCTs reported the role of the sponsor. Of these, 20 explicitly stated that the sponsor had no role in the study. Twenty-one papers described the sponsor's role, the degree of sponsor involvement was variable and usually described vaguely. Among these papers, four stated that researchers had full access to all data, one that the researchers had no limits on publication and one that 'the decision to submit the paper for publication was determined by the study sponsor'. In conclusion, no researcher should be expected to produce 'findings' without full access to the data, freedom from interference in analysis and interpretation and liberty to publish all results, however disappointing to the stakeholder they may be. In the meantime, researchers do well to arm themselves with the rules for research partnerships and editors to take on the role of watchdog.