84 Chapter 4 Sensitivity analyses Using the pre-operative BMI instead of the reference BMI for the study of Villalba et al. resulted in a comparable overall effect of DBS on primary outcome BMI change after DBS, with a large effect size (Hedges’s g = 0 ∙ 90; 95% CI = 0 ∙ 58 to 1 ∙ 21; Z-value = 5 ∙ 60; P < 0 ∙ 001). DISCUSSION This first meta-analysis combined all available trials on the effect of DBS in AN. Four non-randomized non-controlled studies provided a total of 56 participants. Random effects meta-analysis showed an improvement in primary outcome BMI, with a large effect size of 1 ∙ 13 and no heterogeneity. Moreover, meta-analysis showed a beneficial effect on overall psychiatric symptom severity, with a comparable large effect size of 0 ∙ 89. All four symptom domains showed a large effect: depressive symptoms (Hedges’s g = 0 ∙ 98), obsessive-compulsive symptoms (Hedges’s g = 0 ∙ 72), symptoms of anxiety (Hedges’s g = 0 ∙ 85), and eating disorder symptoms (Hedges’s g = 0 ∙ 98). Furthermore, DBS also improved the quality of life, with a large effect size (Hedges’s g = 0 ∙ 86). There was limited evidence for publication bias. The quality of evidence was considered moderate for the more objective primary outcome BMI, and low for the subjective secondary psychological outcome measurements. All in all, results suggest that DBS has statistically large beneficial effects in severe and life-threatening treatment-refractory AN. Although no statistical heterogeneity was found between included studies, some differences are worth mentioning. It is important to note that the study of Villalba et al. had a relatively short period of follow-up of 6 months. This is particularly noteworthy in comparison to the study of Oudijn et al., where an optimization period was included prior to the maintenance phase leading to a maximal post-operative follow-up of 91 weeks. This shorter follow-up may explain the relatively low post-intervention BMI in Villalba et al., although this did not result in statistical heterogeneity between studies. However, the studies used different targets for DBS, possibly resulting in clinical heterogeneity. Another important difference is that only one study reported psychiatric adverse events related to DBS (15). Other studies either experienced no psychological adverse events or did not report them systematically. The risk of bias was considered moderate for objective outcome measurements (BMI) and serious for subjective secondary outcome measurements. This was caused by the lack of a control group, which ethically speaking would be difficult to implement in such a treatment-refractory patient group. A placebo effect could therefore not be excluded. Independent outcome assessors blinded for treatment status could theoretically ameliorate this bias. Nevertheless, we did not find any clear evidence for other biases.
RkJQdWJsaXNoZXIy MjY0ODMw