Stata version 12

All statistical analyses were performed using SPSS 17.0 or STATA version 12. Given that some families contributed more than 1 sibling, hierarchical clustering of data at the level of family was modelled using the multilevel random regression XTMELOGIT or XTMIXED routine in STATA version 12 statistical software [37 ]. Pearson Chi-Square tests, T-test and logistic regression were used when appropriate, in order to determine demographics, illness characteristics and prevalences. Estimates were adjusted for a priori determined confounders (sex, age, ethnicity, marital status (single or not), education level and IQ).

Analysis of efficacy was performed according to the intent-to-treat principle. The odds ratio (OR) was used to quantify the effect of the treatment on tumor response, and its significance was assessed using the Mantel-Haenszel test. The hazard ratio (HR) was estimated to assess the survival advantage of the RT + TMZ as compared with radiotherapy alone. OS and PFS curves were estimated using the Kaplan-Meier methodology and compared using the stratified log-rank test. HR for each group of trials and for all trials together were calculated according to the published methods of Parmar et al. [19 (link)] P values were determined by log-rank test. Toxicity variables were dichotomized as severe (grade 3 to 4) and no/mild (grades 0 to 2). The pooled odds ratio with 95% CI was used to compare toxicity rates between arms.
χ² heterogeneity tests were used to test for statistical heterogeneity among trials. I² statistic was used to estimate the percentage of total variation across studies, with an I² value below 50% considered indicative of low heterogeneity [20 (link)]. If there was a substantial heterogeneity, a random effect model was tested and the possible clinical and methodological reasons for this were explored qualitatively. All tests were two-sided (P = 0.05 was significant). All statistical analyses were performed using STATA version 12.0 software (College Station, TX).

Summary ORs and 95% CIs were calculated to assess the effect of SDF-1 rs1801157 polymorphism on SLE risk under AA vs. GG, AA + GA vs. GG, AA vs. GG + GA, allele A vs. allele G, and GA vs. GG genetic models. Stratified analysis was performed on the basis of ethnicity to further explore the specific correlation between them. The statistical significance of pooled OR was determined by Z test. Deviation of genotype distribution from HWE in control populations was measured using the chi-square goodness-of-fit test. Between-study heterogeneity was checked by chi-square based Q-test. If statistically significant heterogeneity (P < 0.05) existed among included studies, random-effects model was adopted to estimate pooled ORs [26 (link)]; or else, the fixed-effects model was appropriate [27 (link)]. Sensitivity analysis was conducted to detect the reliability of our results by deleting each included study in turn. Publication bias was inspected by Begg’s funnel plot and Egger’s regression test [28 (link)]. All analyses were implemented using STATA version 12.0 software (Stata Corporation, USA). P < 0.05 was considered to be statistically significant.