A total of 103 schizophrenic patients were enrolled from the three collaborating hospitals of Korea University Hospital. All subjects were examined by trained psychiatrists using the Korean version of the Structured Clinical Interview for DSM-IV, leading to a diagnosis based on DSM-IV criteria.²⁴ Exclusion criteria included evidence of other psychiatric, medical, or neurological illness; family history of diabetes or eating disorder; and age over 65 or under 18 years. Application of these criteria resulted in the exclusion of 25 patients. All the subjects were ethnic Koreans, and some findings from these subjects have been reported previously.^25-27 Written informed consents were obtained, and the study protocol was approved by the Ethics Committee of the Korea University Hospital.

Seventy-eight subjects were weighed prior to starting olanzapine and again after long-term treatment at least 3 months (440±288 days). The dosage was adjusted individually according to clinical judgment. We controlled the use of drugs other than olanzapine. Medications such as antipsychotics, mood stabilizers, and antidepressants were avoided during the study, because of their potential effects on weight change. However, we combined the use of benzodiazepines or anticholinergics as needed. No subject had received olanzapine or clozapine prior to the present study. The mean daily dose of olanzapine at the end-point examination was 14.05 mg (standard deviation=5.1 mg).

Other clinical variables that were measured in the study were gender, age, olanzapine treatment duration and dosages, and previous antipsychotics dosages (expressed as chlorpromazine equivalents). Changes in body weight and BMI during the treatment were also calculated.

Genetic polymorphism analyses for the GSTM1 and GSTT1 genes were determined using the multiplex polymerase chain reaction (PCR) with modifications of the previously described method.²⁹ The appropriate fragment of the GST gene for the GSTM1 and GSTT1 alleles was amplified with specific primers from human genomic deoxyribonucleic acid (DNA). The following primers were used in the PCR reaction: GSTM1, (sense) 5'-GAACTCCCTGAAAAGCTAAAGC-3' and (antisense) 5'-GTTGG-GCTCAAATATACGGTGG-3'; GSTT1, (sense) 5'-TTCCTTACTGGTCCTCACATCTC-3' and (antisense) 5'-TCACGGGATCATGGCCAGCA-3'. PCR was performed in a total volume of 30 L containing 100 µg genomic DNA, 5 M of each primer, 2.5 mM deoxyribonucleoside triphosphates, 1.5 mM MgCl₂, 100 mM Tris-HCL, and 1 U thermostable Taq DNA polymerase with a GeneAmp PCR system 2700 (Applied Biosystems, Foster, CA, USA). The amplification conditions were initial denaturation at 94℃ for 5 min, followed by 35 cycles of denaturation at 94℃ for 45 sec, annealing at 59℃ for 50 sec, extension at 72℃ for 1 min, and a final extension step at 72℃ for 10 min. The PCR amplification products (GSTM1: 215 bp, GSTT1: 480 bp) were then separated electrophoretically on an ethidium bromide-stained 2% agarose gel.

A PCR-based assay was used to detect the GSTP1 polymorphism on exon 5.³⁰ The primer sequences for GSTP1 exon 5 were (sense) 5'-GAGGAAACTGAGACCCACTGAG-3' and (antisense) 5'-AGCCCCTTTCTTTGTTCAGCC-3'. A typical PCR reaction was performed in a 25 µL volume containing 1×PCR buffer, 3.0 mM MgCl₂, 0.25 mM dNTPs, 1.5 units of Taq polymerase, and 0.3 mM of primers GSTP1 exon 5. The DNA chains were denatured by incubation at 94℃ for 5 min, followed by 35 cycles of 94℃ for 30 sec, then 60℃ for 30 sec, and 72 ℃ for 30 sec, followed by a final extension step at 72℃ for 5 min. A 424-bp DNA fragment was amplified for exon 5 and followed by 3 h of digestion with 4 units of BsmAI for exon 5 (New England Biolabs, Beverly, MA, USA). The fragments were separated on a 3% agarose gel stained with ethidium bromide. The Ile/Ile, Ile/Val and Val/Val genotypes yielded two bands (292 and 132 bp), four bands (292, 222, 132 and 70 bp), and three bands (222, 132 and 70 bp), respectively.

Differences in allele frequencies between groups with different body weight changes were evaluated by a chi-square analysis and Fisher's exact test. The association of genotype with weight gain and change in BMI was tested with Student's t-test. All of the analyses were performed using standard software (Statistical Package for the Social Sciences for Windows), and p values smaller than 0.05 were considered statistically significant. The multifactor dimensionality reduction (MDR) approach was designed to detect gene-gene interactions in the presence or absence of main effects in case-control studies in human genetics.^31,32 MDR has been shown to have high power for detecting interactions in a wide range of simulated data.^33-35 MDR has also been used to identify interactions in common complex diseases.^36-39 MDR is a non-parametric, model-free approach, making it a unique tool for identifying interactions. MDR categorizes all genetic data into 2 groups, "high risk" and "low risk", by comparing all single loci and all multilocus combinations and then categorizing each genotype as either "high risk" or "low risk" on the basis of the ratio of cases to controls having that genotype. MDR ultimately selects one genetic model, either single or multilocus, that predicts phenotype with the greatest success. To evaluate the predictive ability of the model, prediction error was calculated using 10-fold cross-validation. The result is a set of models, one for each model size considered. From these models, a final model is chosen on the basis of minimization of prediction error and maximization of cross-validation consistency (CVC) (number of times a particular set of factors is identified across the cross-validation subsets). Statistical significance is determined empirically by permuting the case and control labels 1,000 times. The use of permutations to generate p values eliminates the problem of multiple testing. The MDR analysis was carried out using version 1.1.0 of the MDR software package (http://www.epistasis.org).