This study caught my eye recently (here's a link to the abstract:)
It's a naturalistic study, published in BMC Psychiatry in 2010, prospectively following 213 patients with symptoms of psychosis, who were randomized to receive one of four antipsychotic medications (quetiapine, ziprasidone, olanzapine, or risperidone), then apparently followed for up to 2 years.
My prediction with such a study would be that all four medications would have similar effectiveness, with a slight edge in favour of olanzapine.
In fact, the results showed a slight edge in favour of quetiapine. There were no substantial differences in tolerability.
The problems with this study, though, include the following:
1) Most of the data was actually for patients who had only been followed up for 6 weeks (not 2 years!). Only 8 of the 213 patients were followed up for 2 years. Of these 8, 5 were taking olanzapine, 2 were taking ziprasidone, and 1 was taking risperidone. Perhaps one might be tempted to conclude that olanzapine is the drug that has the highest chance of being acceptable for long-term use.
2) The results were presented in a type of "refined" fashion, for example the changes in symptom scores for each drug over time were presented as graphs with a single straight line for each drug, plotted over a 300-day period. This type of graph omits a tremendous amount of relevant data: first of all, there were very few patients who were actually followed for 300 days, most of them were only followed for 6 weeks. A graph like this implies that there are strong data points stretching out over the entire period. Secondly, the linear plots do not show the degree of scatter in the data points. There were no direct reports of the raw data in the study, only refined statistical distillations. It would be much more informative to show all of the data points plotted out over time: then one could see the times where most of the data were derived, the various courses of symptom change for each individual in the study, etc. It would be a messier graph! -- but it would not mislead the viewer to immediately conclude that one medication is obviously better than the others.
In conclusion, the study really grabbed my attention when I first looked at it, but I found it to be much weaker than I thought, after reading it closely.
It does, however, provide a little bit of support for the idea that any one of these four antipsychotic medications are reasonable to try, in the treatment of psychotic symptoms. I agree that quetiapine is a reasonable first choice, though the others could be reasonable also, depending on personal preference, past experience, side effect risks, etc. I would still lean towards olanzapine for anticipated long-term treatment of severe symptoms.