Elizabeth Pisani and colleagues highlight obstacles to data sharing in public health. However, they omit a potentially crucial one: investigators manipulating evidence that might contradict their hypotheses or conclusions, as illustrated by the recent scandal of some climatologists allegedly withholding data for global temperatures.
We have freely shared sensitive, deidentified data on HIV epidemiology for nearly two decades. Even with colleagues in poor countries, we have archived such data publicly. If a dataset can be analysed and results published, it takes little effort to make a suitable dataset available to other researchers, because nearly all the necessary data-management work has already been done.
We and others have made numerous requests for similar de-identified data from British and American investigators studying HIV transmission in sub-Saharan Africa, including one team involving a coauthor of Pisani and colleagues' current Comment. In all instances, the requests were denied, even when the journals that published the investigators' results had policies mandating data sharing.
When research data are not shared, the reported results could be considered suspect, especially when the paradigm underlying the research is contested. Indeed, publicly available data from the Demographic and Health Surveys on HIV epidemiology in sub-Saharan Africa undermine the assertion that HIV in that region is transmitted almost exclusively through sex and from mother to child. The public trust in and support of science depend crucially on researchers sharing data, not merely advocating such sharing.
We declare that we have no conflicts of interest.
Link to access full article: http://www.thelancet.com/journals/lancet/article/PIIS0140-6736(10)60870-7/fulltext