This varies greatly. I have seen several studies showing differences in efficacy based on demographic. There seems to be fantastic outcomes for at-risk secondary students. However, for a lot of other groups, online learning just does not work.
I have not seen studies based on subject, but I am sure they exist. I would assume that subject and course will vary greatly. I would not mind taking financial accounting online. However, computer science 101 would be very difficult for me. Accounting seems to be taught ok by email and spreadsheets. Mistakes can be communicated in a mostly satisfactory way. However, in programming, conceptual vs. syntactical bugs can prove difficult to communicate. This difficulty in communication can lead to misconceptions.
Student motivation plays a key part in anything education related.
Efficacy of online learning depends on too many factors. Basically, it is very effective when provided as a choice in certain subjects when a particular demographic is present and satisfactory tools are available.