Given that the publication process is sometimes slow, this does occur. In principle, I don't think there is anything wrong with it. In my own experience as the statistician/data analyst, I have certainly been asked questions of reviewers that have led to new or additional analyses as part of the revisions (I was comparatively willing to do that--as a reviewer, I find most authors pretty resistant). I don't recall dramatically different findings, but the risk would be there. In a voluntary re-analysis of my own data as part of a methods article, I found that under new conditions, the effect was only half as big as originally reported, but it might have been the case that there was no longer an effect, for example. So the risk one takes is that by the time that under revision article makes it to press the result used as the basis for another study may not hold and no longer can be referenced to that study. So in making a decision to base a study on an under revision study would rest on how important it might be that the result in that study could change in the published version. Of course, if one is the author of that study or in touch with the authors, one would know if the results themselves, as opposed to other aspects of the article, are questioned. So, if not, there shouldn't be much risk. Of course, under revision doesn't guarantee publication, so if the credibility of publication is important, that isn't there yet. Bob
It is better to use the highly cited papers or the article which authored by a scientist which has highest H-index. https://www.academia.edu/1738979/How_to_Conduct_a_Literature_Review
Since most journals employ a blind review process, citations of unpublished work can give away the author of the paper, thus potentially biasing the review process. It is up to the author to decide if he wants that risk or not. APA guidelines, however, suggest that one should avoid citing unpublished work.