In addition to non-experimentally obtained or non-replicated results, anecdotal evidence applies similarly to scientific experimental results. The same evidence can be considered both anecdotal and well-supported, based on the degree of scrutiny that the evidence has passed.
For example, if I perform a number of experiments in my lab, obtain a preliminary result from the experiments, and then tell you that I had obtained a certain result, the evidence would be considered anecdotal. This is because I could have committed one of several mistakes within the experimental construction, statistical analysis, result interpretation, or many other error sources that plague scientific research.
However, if I perform rigorous statistical analysis on the data, and show that the data is statistically significant and has a large effect size, and the experiment goes through peer review and is published in a journal, this evidence becomes well-supported. This is because the process of peer review checks for these sources of error within the experiment, and is likely to prevent publication if significant errors are present. While peer review is obviously not foolproof, it removes a significant amount of experimental conclusions which are not valid.
Therefore, a citation from a paper that is peer reviewed should be rated as far more valid and provide far stronger backing for an argument than anecdotal evidence.