Thirty-two case reports (n=32) and 13 case series (n=284) were included.
Flaws included potential for reporting bias, lack of control group, lack of diagnostic criteria, lack of independent outcome assessment or use of standardised scale and inadequate follow-up.
Case reports: the mean GIS score was 2.81 out of 7 (range 1 to 7) suggesting a small improvement at follow-up. Seven patients scored 4 or more suggesting no change or a worsening of symptoms; the two patients who scored 7 committed suicide. There was no significant difference in GIS scores between patients treated with confrontation versus non-confrontational techniques, single versus multiple treatments, interventions that included psychotherapy versus non-psychotherapy interventions, psychiatric medication versus no medication, longer versus shorter treatments or initial treatment as inpatient versus outpatient.
Case series: the median GIS score was 0 due to lack of follow-up of most patients. Four patients (3.6 per cent) died. For studies reporting follow-up on more than 80 per cent of patients (seven studies, n=40) median GIS was 2, suggesting improvement.