Our findings suggest that G-CSF can sometimes be safely used to support the maintenance of normal neutrophil counts and clozapine use after neutropenia. Publication bias is an important limitation, however. Also, few reports clearly documented the presence or absence of an independent nonclozapine cause of the index neutropenia, which may have increased success rates. Furthermore, adverse events were not systematically recorded. Prospective studies are needed to determine safety because if agranulocytosis occurs on clozapine while supported by G-CSF, there is no obvious alternate rescue therapy to promote granulopoiesis. From the available data, it is not possible to recommend this course of action for someone with a true clozapine agranulocytosis.

Our findings indicate that G-CSF/GM-CSF use is well tolerated and suggest that G-CSF can sometimes be safely used to reduce the duration of neutropenia associated with clozapine use. However, the interpretation of this outcome is difficult, given the likely publication bias for positive outcomes in case reports.

We identified an association between the concurrent use of sodium valproate and an increased risk of clozapine associated neutropenia. These results, taken in combination with the results from previous case series, suggest that the risk of clozapine associated neutropenia could be reduced by avoiding concurrent valproate treatment.

Antibodies to the voltage-gated potassium channel (VGKC) complex and glutamic acid decarboxylase (GAD) have been reported in some cases of psychosis. We conducted the first systematic review and meta-analysis to investigate their prevalence in people with psychosis and report a case series of VGKC-complex antibodies in refractory psychosis. Only five studies presenting prevalence rates of VGKC seropositivity in psychosis were identified, in addition to our case series, with an overall prevalence of 1.5% (25/1720) compared to 0.7% in healthy controls (12/1753). Meta-analysis established that the pooled prevalence of GAD65 autoantibodies was 5.8% (95% confidence interval [CI]: 2.0-15.6%; I = 91%; nine studies) in psychotic disorders, with a prevalence of 4.6% (95%CI: 1.2-15.9%; nine studies; I = 89%) and 6.2% (95%CI: 1.2-27.0%; two studies; I = 69%) in schizophrenia and bipolar disorder, respectively. People with psychosis were more likely to have GAD65 antibodies than controls (odds ratio [OR], 2.24; 95%CI: 1.28-3.92%; P = 0.005; eight studies; I = 0%). Among 21 participants with treatment-resistant psychosis, none had VGKC antibodies. The prevalence of VGKC antibodies is low in psychosis. Our preliminary meta-analysis suggests that GAD autoantibodies are more common in people with psychosis than in controls, although few studies accounted for the possibility of co-existing type 1 diabetes mellitus and the clinical significance of reported GAD titers remains unclear. The paucity of studies reporting thresholds for defining GAD abnormality and rates of comorbid type 1 diabetes mellitus precludes interpretations regarding the influence of GAD antibodies on the development of psychotic disorders and may have led to an overestimate of the prevalence of GAD. Our case series fails to support the hypothesis that VGKC antibodies are linked to treatment resistance in psychosis, but the literature to date is remarkably sparse.