There are a number of limitations to all of the research studies published to date. For example
The systematic review by Mercer (Epub) provides a useful analysis of some of the other limitations of these studies.
“These studies share a number of limitations. One important point is that they have not systematically compared DIR/Floor-time to a treatment for which there is clear evidentiary support of effectiveness. The only such treatments are discrete trial methods, and although many of the children in both intervention and control groups were having some ABA treatment, the published reports do not include any information that enables readers to compare ABA to DIR as might be considered appropriate, given the empirical support for ABA. A second problem is that the studies are designed to compare DIR plus usual services to usual services alone; in other words, each DIR or DIR-like group receives many more hours of treatment than the control group does, raising the obvious question whether any positive effects are simply due to more interaction with adults rather than with DIR specifics. (This problem is present even in the sophisticated and carefully implemented Solomon et al., 2014, study.)
“Additional limitations of these studies are very difficult to avoid and emerge from the difficulty of studying families as they function in their homes. Families dropped out of studies or changed their practices with their children as a result of learning what other parents were doing. How many hours were actually spent in Floortime could be known only from parent reports. Other services received by either the intervention or the control groups were not necessarily known, and if reported at the beginning of a study could well have changed by the end. Casenhiser et al. (2013) also noted that in their study there were unavoidable self-selection biases, as the parents involved had to be amenable to DIR methods, had to be able to attend training 2 hours a week, had to be able to complete assignments, and had to be able to spend at least 3 hours each day in interactions with one child. Given the difficulties of this type of research, it seems reasonable to conclude that the existing studies give weak support to the effect of Floortime on some skills of autistic children, although improved language functioning is not among these. However, no independent replications of any of the studies have been carried out.”
For a comprehensive list of potential flaws in research studies, please see "Why some autism research studies are flawed."