Strengths, limitations and interpretation
In addition to earlier published reviews concerning the association between endometriosis and psychological problems [21,96], the current review specifically provides an up to date overview with regard to both depressive and anxiety symptoms in these patients. Furthermore, we provided information concerning factors that correlate with depression and anxiety symptoms in endometriosis. Reporting of these correlations is important to recognise vulnerable patients in clinical practice. As research demonstrated the reliability of non-surgical diagnosis of endometriosis, we also included studies including patients with a clinical diagnosis of endometriosis, which made it possible to generate a large number of studies. In contrast to Gambadauro et al, our review excluded studies with self-reported diagnoses of depression and studies using non standardized assessment methods, to limit the risk of over-reporting. The meta-analysis included solely studies using a standardized outcome measure to limit heterogeneity in the results. The overall SMD of studies comparing endometriosis patients with CPP patients without endometriosis showed an I-squared of 0% (p=0.798) for depression scores and an I-squared of 33.2% (p=0.134) for anxiety scores, which means that the heterogeneity between these studies was low. Nevertheless, heterogeneity between studies comparing endometriosis patients with healthy controls was relatively high: I-squared 84.6% (p<0.001) for depression scores and I-squared 68.1% (p<0.001) for anxiety scores. This makes it difficult to compare and generalize the study outcomes. Hence, strategies of research groups and their assessment tools need to be standardized to produce robust and reliable results for meta-analyses. However, a sub-analysis including only studies using the HADS as outcome measures, did not change outcome nor improved I-squared results, meaning that heterogeneity was probably caused by differences in demographic or other factors. Furthermore, this review has some other limitations. First, including articles in English only limits the number of included studies. Risks of bias were mostly caused by study design and patient selection as most studies included women in tertiary referral centres. This could have led to an over-presentation of symptoms compared to patients in the general population. Although validated, the used assessment tools are predominantly self-assessment questionnaires, which gives an indication of a depressive or anxious state but does not provide a valid diagnosis. Studies using reported ICD-9 diagnoses given by professionals could have led to under-reporting of depression and anxiety complaints as those might have not received an official diagnosis. Another important limitation is the variation in types of anxiety disorders that were pooled in the meta-analysis. Screening tools for anxiety disorders usually provide lower sensitivity and specificity than tools for depressive disorders, as anxiety disorders have more heterogeneous symptoms (state, trait or phobic anxiety) [94].