Studies that report that a treatment works are considered more interesting than studies that report that a treatment doesn’t work. As a result, studies that report that a treatment works are more likely to get published, leading to a skewed perception of the ratio of positive to negative studies.

To take an extreme example, let’s say that 10 equally good studies about some intervention were performed. One of them had positive results (the treatment was found beneficial) and 9 of them had negative results (the treatment was found to have no effect, or a negative effect). But if the single positive study got published and the 9 negative studies didn't, it would look like the treatment was beneficial, when in fact most of the evidence suggested that it was not.