Ity of reporting, making it possible to extract the relevant information. Despite the known weaknesses of meta-analysis, the large sample and the homogeneity of the results after adjustment add weight to our conclusions: we found that FGFR3 and TP53 mutations occurred independently 
when stage and grade were taken into account, and that the CAL 120 site frequency of TP53 mutation was high in pT1G3 and pT2-4 tumours, regardless of the presence or absence of FGFR3 mutations in these tumours. Thus, TP53 mutations can occur 1676428 in Ta pathway tumours, when these tumours progress. However, the time frame of TP53 mutation differs considerably between the two pathways: TP53 mutations occur before basement membrane invasion in the carcinoma in situ pathway, whereas they probably occur after or during basement membrane invasion in the Ta pathway. Our findings also indicate that care is required in the analysis of bladder tumours, as both pathways of tumour progression for bladder cancers 15481974 must be taken into account when interpreting data.stages (pT1-4) (Ouerhani et al., 2009) or at any stage. Three studies concerned newly diagnosed patients (Bakkar 2003, Hernandez et al., 2005; Lamy 2006). Two of these studies have not yet been published and are denoted below by the name of the main investigator or the program supporting the study plus “UP” (for unpublished; thus, Mongiat-Artus UP and BladderCIT UP). Information about treatment (irradiation or chemotherapy) before Hexokinase II Inhibitor II, 3-BP site biopsy was unavailable for a large proportion of the patients. We therefore included all patients, regardless of their prior treatment. The main characteristics of these eight studies are summarised in Table 1. We refer to the publications for all details in the cases of published data. In the Bladder CIT UP study, the pTa and pT1 tumours were from incident cases and the pT2-4 tumours were from both newly diagnosed and progressing cases (patients with a history of previous non-muscle-invasive tumours). FGFR3 mutations were assessed by the SNaPshot technique in the BladderCIT UP study [28] and by allele-specific PCR [29] in the Mongiat-Artus UP study. TP53 mutations were assessed by direct sequencing on both strands, followed by confirmation of the identified mutations in an independent PCR in the BladderCIT UP study and by the FASAY method [30,31] followed by confirmation of the identified mutations by an independent PCR in the Mongiat-Artus UP study. We did not include the study by van Rhijn et al. (2004) [32] because TP53 alterations were assessed by immunohistochemistry in this study and it has been shown that there is only 57 to 71 similarity between the alterations detected by mutation assessment and those detected by immunohistochemistry [Table S3]. The following variables were collected: TP53 and FGFR3 mutations, stage and grade of the disease and the type of mutation. Data for individual patients were available for four studies (Mongiat-Artus UP, BladderCIT UP, Lindgren et al., 2006, Ouerhani et al., 2009) [Table S4 and S5]. For the other reports, data were extracted directly from the publications by crossreferencing tables. For three studies (Bakkar et al., 2003; Lindgren et al., 2008; Ouerhani et al., 2009), the authors were contacted and provided additional information.Biological and pathological dataThe frequencies of FGFR3 and TP53 mutations by grade and tumour stage were recorded in all studies. Data were extracted from the publications or from the raw data, when available [Table S4 and S.Ity of reporting, making it possible to extract the relevant information. Despite the known weaknesses of meta-analysis, the large sample and the homogeneity of the results after adjustment add weight to our conclusions: we found that FGFR3 and TP53 mutations occurred independently when stage and grade were taken into account, and that the frequency of TP53 mutation was high in pT1G3 and pT2-4 tumours, regardless of the presence or absence of FGFR3 mutations in these tumours. Thus, TP53 mutations can occur 1676428 in Ta pathway tumours, when these tumours progress. However, the time frame of TP53 mutation differs considerably between the two pathways: TP53 mutations occur before basement membrane invasion in the carcinoma in situ pathway, whereas they probably occur after or during basement membrane invasion in the Ta pathway. Our findings also indicate that care is required in the analysis of bladder tumours, as both pathways of tumour progression for bladder cancers 15481974 must be taken into account when interpreting data.stages (pT1-4) (Ouerhani et al., 2009) or at any stage. Three studies concerned newly diagnosed patients (Bakkar 2003, Hernandez et al., 2005; Lamy 2006). Two of these studies have not yet been published and are denoted below by the name of the main investigator or the program supporting the study plus “UP” (for unpublished; thus, Mongiat-Artus UP and BladderCIT UP). Information about treatment (irradiation or chemotherapy) before biopsy was unavailable for a large proportion of the patients. We therefore included all patients, regardless of their prior treatment. The main characteristics of these eight studies are summarised in Table 1. We refer to the publications for all details in the cases of published data. In the Bladder CIT UP study, the pTa and pT1 tumours were from incident cases and the pT2-4 tumours were from both newly diagnosed and progressing cases (patients with a history of previous non-muscle-invasive tumours). FGFR3 mutations were assessed by the SNaPshot technique in the BladderCIT UP study [28] and by allele-specific PCR [29] in the Mongiat-Artus UP study. TP53 mutations were assessed by direct sequencing on both strands, followed by confirmation of the identified mutations in an independent PCR in the BladderCIT UP study and by the FASAY method [30,31] followed by confirmation of the identified mutations by an independent PCR in the Mongiat-Artus UP study. We did not include the study by van Rhijn et al. (2004) [32] because TP53 alterations were assessed by immunohistochemistry in this study and it has been shown that there is only 57 to 71 similarity between the alterations detected by mutation assessment and those detected by immunohistochemistry [Table S3]. The following variables were collected: TP53 and FGFR3 mutations, stage and grade 
of the disease and the type of mutation. Data for individual patients were available for four studies (Mongiat-Artus UP, BladderCIT UP, Lindgren et al., 2006, Ouerhani et al., 2009) [Table S4 and S5]. For the other reports, data were extracted directly from the publications by crossreferencing tables. For three studies (Bakkar et al., 2003; Lindgren et al., 2008; Ouerhani et al., 2009), the authors were contacted and provided additional information.Biological and pathological dataThe frequencies of FGFR3 and TP53 mutations by grade and tumour stage were recorded in all studies. Data were extracted from the publications or from the raw data, when available [Table S4 and S.
