Study Identification and Selection. Where relevant to the present review the criteria set out in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist were followed (Moher, Xxxxxxxx et al. 2009). Excluded (including duplicates): (n=639) Total records: (n=649) Excluded (see text): (n=2) Articles screened: (n=10) Included in analysis: (n=8) Figure 1-5 Flow diagram illustrating results of electronic and manual searches and study identification process From (Xxxxxx, Xxxxxxx et al. 2013) The results of the following identification process and study selection are illustrated in Figure 1-5. A search of PubMed/XXXXXXX and Xxxxxx was conducted using the search term “(dementia or alzheimer*) AND (serotonin or 5ht*) AND (polymorphism or gene*)”. The only limit on the search was that articles had to be written in English. The resulting 649 records over the two lists were screened for candidate gene association studies of the 5HTTLPR polymorphism in relation to psychotic, delusional or hallucination symptoms in Alzheimer’s disease. References contained in review articles and meta-analyses (XxXxxxxxx-Xxxxx, Xxxxx 2010, Xxxxxxx, Xxxxxxxx et al. 2010, Xxxxxxxxxx, Xxxxx 2009, Xxxxxxx, Sobow et al. 2011) were also manually searched for any additional studies not picked up by the above search. Any duplicate articles were excluded. Ten articles were found using the aforementioned method and these were screened in detail to meet the following inclusion criteria: 1) standardised diagnosis of Alzheimer’s disease; 2) psychotic symptoms (delusions, hallucinations or psychosis) assessed using standardised assessment tool; 3) level of cognitive impairment measured using MMSE and reported. One study (Xxxxxx, Xxxxxxx et al. 2003) did not report MMSE, leaving nine in total. However, two were carried out by Xxxxxxx et al. in 2006 (Xxxxxxx, Xxxxxx et al. 2006, Xxxxxxx, Xxxxxx et al. 2006), using very similar cohorts, both reporting significant findings. In order to exclude any bias resulting from including non-independent cohorts, the Borroni et al. study with the analysis most similar to the other studies was included, that is the one that examined symptom frequencies using logistic regression rather than factor analysis (Xxxxxxx, Xxxxxx et al. 2006), leaving eight studies in the final review (Table 1-1).
