[AnotherHustler]

Additionally, the study does not differentiate between therapy types received - treatment modalities vary in efficacy depending on person and context. For example, 20 years ago it was known within psychology that "Cognitive Therapy" had about a 50/50 success rate. However, if you try to apply cognitive therapy to someone diagnosed with borderline personality disorder - it's not likely to work ever - and might actually make things worse (Dialectic Behavioural Therapy is probably the gold standard for this).

As you travel down the road towards being a psych. professional - you tend to subscribe a school of thought about behavior and personality, and there are a few to choose from (e.g. people are generally self-directed and will solve their own problems given the opportunity, or behavioural problems have their roots in learning - so training is required) - and depending on your opinion you'll eventually subscribe to a treatment modality that resonates with what you believe (the extreme version of this is abandoning psychology & switching to medicine / psychiatry because you believe behavior is simply a neurochemical process).

Serious personality disorders are very difficult to treat, which brings me to my next point. The Big-Five factor model used to determine change over time does not factor diagnosis, so we don't really get a picture of what the cohort looks like. We know nothing about patient history or reason for seeking therapy (i.e. have you been evaluated admitted to a psychiatric ward in the past? have you suffered from substance abuse problems?). And finally, the big question that would be useful is "what problems caused you to seek therapy?" Depression vs. hearing voices in one's head vs. got divorced because I like to drink are all very different issues.

I would like to have seen at least three more measures: reason for treatment, length of treatment and therapy modality used.

However, I will say that stating that "new research proves that therapy doesn't work" will get you a click or two ;-)