SIR—It has been recently reported that the 50 end of Neuregulin 1 gene (NRG1) is associated with schizophrenia in Icelandic and Scottish populations. 1, 2 We confirmed this finding in the present case–control study in a Chinese sample, which analyzed 13 microsatellites extending about 360 kb from upstream of NRG1 to the first intron. The subjects consisted of 540 unrelated schizophrenic in-patients and 279 controls. All case samples were collected from the Shanghai Mental Health Center. Clinical diagnosis was made according to DSM-IIIR criteria by two independent clinicians. A standard informed consent, which was reviewed and approved by the Shanghai Ethical Committee of Human Genetic Resources, was granted by all the participants. Control individuals were from attendees at hematology clinics in Shanghai. From the polymorphisms used by Stefansson et al1 for genotyping, we chose 13 microsatellite markers, which span around an 800kb region of chromosome 8pter. The genotyping assay was performed on MegaBACE 1000 instruments (Amersham Biosciences). Genotype distributions were within Hardy–Weinberg equilibrium in both patients and controls for all the markers except D8S1769, which was then excluded in the later association analysis. Differences in allelic and haplotypic distributions were estimated using the program CLUMP 1.6, 3 with more than 10000 simulations. Statistically significant differences in allele distributions were observed between schizophrenics and controls on four microsatellite markers (Table 1). According to the standardized pairwise disequilibrium D04 values, the 13 markers can be divided into four regions: the first two markers (D8S1770 and D8S1769); the four markers upstream of NRG1 gene (29H12-7320 through 478B14-642); the five markers in the first intron (487-2 through 420M9-3663); and the last two markers (473C15-439 and 72H22-1)(Table 1). In general, the D0 values greater than 0.3 were observed between the markers within the same regions, while hardly between site pairs from different regions. Calculations of the significance of LD revealed a similar pattern. As a result of this LD pattern, haplotype analysis was performed in each region separately. PHASE5 software was applied to estimate the multiple marker haplotype frequencies, then haplotypes with an estimated frequency greater than 1% were included for the χ2 test using CLUMP. 3 For regions 2 and 3, haplotypes with probabilities greater than 1% accounted for the vast majority of the haplotype diversity (82 and 80%, respectively), and overall frequency comparisons showed significant differences between cases and controls (T2, X2¼57. 80, 19 df, P¼0. 00003; T2, X2¼49. 23, 19 df, P¼0. 000267, respectively). The statistic T3 produced by CLUMP for region 2 revealed that the difference reflected mainly different distributions of haplotype ‘0/2/0/2’(frequency¼1. 45 and 4.4%, for case and control group, separately; T3, X2¼13. 80, 1 df, P¼0. 0036; OR¼0. 31, 95% CI: 0.16–0.59). For region 3, the greatest contribution to the χ2 total came from the haplotype ‘20/4/–2/18/0’(frequency¼35. 8 and 27.2%, for case and control group, separately; T3, X2¼11. 80, 1 df, P¼0. 014; OR¼1. 49, 95% CI: 1.19–1.87). For region 4, the two-marker haplotype of 473C15-439 and 72H22-1 yielded no significant P-values (Table 1). Since the microsatellite markers 478B14-642, 487-2, 478B14-848, and 420M9-1395 overlapped the region of the core at-risk haplotype identified in the Icelandic and Scottish samples, we also analyzed the haplotypes of these four markers, finding the P-values were above the significant level (T2, X2¼15. 91, 19 df, P¼0. 6632 …