Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

An Integrated Meta-Analysis of Two Variants in HOXA1/HOXB1 and Their Effect on the Risk of Autism Spectrum Disorders

  • Ran-Ran Song ,

    Contributed equally to this work with: Ran-Ran Song, Li Zou

    Affiliation Department of Maternal and Child Health, School of Public Health, Tongji Medical College, Huangzhong University of Science and Technology, Wuhan, China

  • Li Zou ,

    Contributed equally to this work with: Ran-Ran Song, Li Zou

    Affiliation Department of Epidemiology and Biostatistics and MOE Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China

  • Rong Zhong,

    Affiliation Department of Epidemiology and Biostatistics and MOE Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China

  • Xia-Wen Zheng,

    Affiliation Department of Epidemiology and Biostatistics and MOE Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China

  • Bei-Bei Zhu,

    Affiliation Department of Epidemiology and Biostatistics and MOE Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China

  • Wei Chen,

    Affiliation Department of Epidemiology and Biostatistics and MOE Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China

  • Li Liu ,

    miaoxp@mail.hust.edu.cn (XM); pupuliu919@gmail.com (LL)

    Affiliation Guangdong Key Lab of Molecular Epidemiology and Department of Epidemiology and Biostatistics, School of Public Health, Guangdong Pharmaceutical University, Guangzhou, China

  • Xiao-Ping Miao

    miaoxp@mail.hust.edu.cn (XM); pupuliu919@gmail.com (LL)

    Affiliation Department of Epidemiology and Biostatistics and MOE Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China

An Integrated Meta-Analysis of Two Variants in HOXA1/HOXB1 and Their Effect on the Risk of Autism Spectrum Disorders

  • Ran-Ran Song, 
  • Li Zou, 
  • Rong Zhong, 
  • Xia-Wen Zheng, 
  • Bei-Bei Zhu, 
  • Wei Chen, 
  • Li Liu, 
  • Xiao-Ping Miao
PLOS
x

Abstract

Background

HOXA1 and HOXB1 have been strongly posed as candidate genes for autism spectrum disorders (ASD) given their important role in the development of hindbrain. The A218G (rs10951154) in HOXA1 and the insertion variant in HOXB1 (nINS/INS, rs72338773) were of special interest for ASD but with inconclusive results. Thus, we conducted a meta-analysis integrating case-control and transmission/disequilibrium test (TDT) studies to clearly discern the effect of these two variants in ASD.

Methods and Findings

Multiple electronic databases were searched to identify studies assessing the A218G and/or nINS/INS variant in ASD. Data from case-control and TDT studies were analyzed in an allelic model using the Catmap software. A total of 10 and 7 reports were found to be eligible for meta-analyses of A218G and nINS/INS variant, respectively. In overall meta-analysis, the pooled OR for the 218G allele and the INS allele was 0.97 (95% CI = 0.76-1.25, Pheterogeneity = 0.029) and 1.14 (95% CI = 0.97-1.33, Pheterogeneity = 0.269), respectively. No significant association was also identified between these two variants and ASD risk in stratified analysis. Further, cumulative meta-analysis in chronologic order showed the inclination toward null-significant association for both variants with continual adding studies. Additionally, although the between-study heterogeneity regarding the A218G is not explained by study design, ethnicity, and sample size, the sensitive analysis indicated the stability of the result.

Conclusions

This meta-analysis suggests the HOXA1 A218G and HOXB1 nINS/INS variants may not contribute significantly to ASD risk.

Introduction

Autism spectrum disorders (ASD) comprises a pervasive group of neurodevelopmental disorders that share common features of impaired social interaction, deviance in language development, and stereotyped behaviors or narrow range of interests [1]. The prevalence rate of ASD is 0.6%–1% in child and adolescent populations, making it one of the most common disorders of development in the world [2], [3]. The etiology of ASD has been debated ever since, but twin and family studies have highlighted the genetic contribution to ASD, with a hereditability as high as 90% [4]. Converging lines of evidence pointed toward altered prenatal neurodevelopment as being crucial to ASD pathogenesis, suggesting a failure of development of at least one of the rhombomeres, from which facial nucleus and superior olive arose [4]. In this regard, one relied on available biological evidence that genes encoding proteins involved in early neural development could thus underlie this disease [4].

HOXA1 and HOXB1, located on chromosomes 7p and 17q, are two paralogous genes in the HOX gene family of homeobox transcription factors and critically involved in the developing hindbrain during neural tube formation [5]. Evidence from autopsy on brain of an idiopathic autism patient has showed the near absence of the facial nucleus and superior olive [6], while the facial nucleus and superior olive required normal HOXA1 and HOXB1 function for their proper development. Similar morphologic deficits in brain have been described in mice knockout Hoxa1 and Hoxb1 [7]. Moreover, Hoxa1 and Hoxb1 are highly expressed during an early period of development overlapping with the window of maximal prenatal sensitivity in rodent model [8]. These evidence strongly posed HOXA1 and HOXB1 as potential candidate genes for ASD. Ingram and his collaborators firstly described variants in HOX genes for genetic susceptibility of ASD, including an A to G substitution at base 218 (A218G, rs10951154) of HOXA1 and a 9-base insertion variant (nINS/INS, rs72338773) of HOXB1 [9]. Of these two candidate variants, the A218G in HOXA1 changes a histidine (H) to arginine (R) at position 73 (H73R) and disrupted a string of histidine repeats in exon 1, while the 9-base insertion c.82insACAGCGCCC (referred as INS allele) in exon 1 of HOXB1 introduces into the amino acid sequence the tripeptide histidine-serine-alanine (H-S-A) [10]. Ingram et al. firstly reported significant deviations from Hardy-Weinberg equilibrium for A218G of HOXA in 57 autistic probands and 119 unrelated adults as controls, indicating that the G allele carriers might be susceptible to ASD [9]. Interestingly, interaction among HOXA1, HOXB1, and gender was also found to be associated with increased risk of ASD [9]. After this first report, a string of studies have subsequently detected these two variants of HOXA1 and HOXB1 in ASD, but the results were inconsistent. Inversely, Conciatori et al. pointed out the A allele, not the G allele, was associated with ASD risk in both case-control and family-based association analyses [11]. Additionally, some researches failed to replicate the findings by Ingram et al. [10], [12], [13], [14], [15], [16], [17], [18], [19]. The sample size of any individual study tended to be small and thus potentially leaded to imprecise estimates and inconsistent results for these two variants in ASD. Nevertheless, meta-analysis, due to its exponential increase in sample size, may be a powerful tool to clarify the inconsistent findings in genetic association studies by statistical synthesis of data if properly used [20]. Furthermore, the transmission/disequilibrium test (TDT) based on family is particularly advantageous as less confounding caused by population admixture and is of the same importance as case-control study in genetic association analysis [21]. However, it might be a little statistically challenging to synthesis the family-based studies with case-control studies before Kazeem et al. outlined a methodological improvement for achieving integration of these two different types of studies by a fixed-effects approach, using an allele-based mode [22], and then Nicodemus subsequently extended this method to the random-effects model [23]. Therefore, we conducted a meta-analysis of published studies and applied the method described by Kazeem et al. [22] to integrate the results from case-control and TDT studies to provide more precise evaluation for the association of HOXA1 and HOXB1 genetic variants with ASD risk.

Methods

To ensure the rigour of this current meta-analysis, we designed and reported it according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) [24] statement and the checklist is shown in table S1 (http://www.prisma-statement.org).

Search strategy and identification of relevant studies

We searched PubMed, EMBASE, and ISI Web of Science databases for published articles up to August 2010, which had investigated at least one of A218G in HOXA1 and insertion variant in HOXB1 associated with ASD in case-control study or TDT study. The search strategy was based on combinations of the keywords ‘HOXA1, homeobox A1, HOXB1 or homeobox B1’, ‘polymorphism or variant’ and ‘autism or autistic disorder’ without language restrictions. To expand the coverage of our searches, we further performed searches in Chinese Biomedical (CBM) database using the above searching strategy. References of the retrieved articles were also scanned. Reviews, comments, and letters were also checked for additional studies.

The inclusion criteria had to be fulfilled: (1) either case-control or TDT study design; (2) data on any or both polymorphisms of A218G in HOXA1 and insertion variant in HOXB1; (3) presentation of data necessary for calculating odds ratios (ORs); (4) clear definition of ASD. Animal studies, reviews, simply commentaries, case reports and unpublished reports were excluded. Study overlapping with other studies should be eliminated, and the report with complete design or larger sample size was finally selected.

Data extraction

All data were extracted independently by two reviewers (R–R. Song & L. Zou). The following information was extracted from the eligible studies: first author's name, year of publication, ethnicity, design type of study, and diagnostic criteria for ASD. Counts of alleles in case and control group in case-control studies and numbers of transmitted alleles from heterozygous parents to affected offspring in family-based studies were extracted or calculated from published data in the included studies.

Statistical analysis

Data from case-control studies were summarized in two-by-two tables, and data from TDT studies were summarized in two-by-one tables. ORs and their 95% confidence intervals (CIs) and standard errors (SEs) were calculated for individual study based on the allele data using the method as described by Kazeem et al. [22]. The χ2- based Cochran's Q statistic test was employed to test between-study heterogeneity, and heterogeneity was considered significant when P<0.1 for Q statistic. Data from the studies were combined by the random-effects model when heterogeneity was significantly present; otherwise, fixed-effects model was applied. For the synthesis of case-control and TDT studies, the method described by Kazeem et al. was used [22]. After obtain the estimate of logarithm of the OR and its associated SE in each case-control or TDT study, the estimate of combined OR and its associated SE can be calculated by a weighted analysis method [25]. The Catmap software implemented this method for the fix-effects model and plus extended this method for the random-effects model of DerSimonian and Laird to conduct case-control and TDT meta-analysis, which can be download from the comprehensive R network http://www.r-project.org [23]. Overall meta-analysis was initially performed. Then stratified analysis, if feasible, was performed according to study design, sample size, and ethnicity separately. Publication bias was assessed by funnel plot and Egger's test [26]. Additionally, sensitivity analysis was also performed to assess the influence of each study on the overall estimate. Cumulative meta-analysis was also conducted via the assortment of studies by publication time. All P values are two-tailed with a significant level at 0.05. All statistical analyses were carried out in Catmap software V1.6.

Results

Characteristics of included studies

Figure 1 shows the literature search and study selection procedures. After comprehensive searching, 23 potentially relevant reports were retrieved, of which, 12 reports met the inclusion criteria. However, the study reported by Muscarella et al. in 2007 [16] was excluded since the cases largely overlapped with the sample in analysis by Conciatori et al. [11]. A total of 11 reports were ultimately eligible for this meta-analysis [9], [10], [11], [12], [13], [14], [15], [17], [18], [19], [27]. Additionally, since 5 reports applied a double approach comprising case-control and TDT designs in the same or overlapping probands [9], [11], [15], [19], [27], the case-control studies of these 5 reports were excluded in the overall meta-analysis due to less sample size than TDT studies. Therefore, 10 reports comprising 113 cases, 184 controls and 958 families were relevant to A218G in HOXA1, and 7 reports including 113 cases, 184 controls and 570 families were relevant to insertion variant in HOXB1. Table 1 shows the characteristics of the included studies.

Combining results of case-control and TDT studies

Figure 2A shows the combined result of case-control and TDT studies for A218G in association with ASD. Heterogeneity was found among the 3 case-control and 10 TDT studies from 10 reports (χ2 = 22.83, Pheterogeneity = 0.029); thus, the random-effects model was employed. Pooled OR and 95% CI were calculated for the 218G allele versus the 218A allele. However, no significant association was found between the allelic variant and ASD risk (OR = 0.97, 95%CI = 0.76-1.25, P = 0.843). For HOXB1 nINS/INS variant, no evidence of heterogeneity was presented among the 3 case-control and 5 TDT studies from 7 reports (χ2 = 8.78, Pheterogeneity = 0.269). In the fixed-effects model, this insertion variant was not significantly associated with risk of ASD (OR = 1.36, 95%CI = 0.97-1.33, P = 0.118; Figure 2B).

thumbnail
Figure 2. Summary estimates (Odds ratios and 95% confidence intervals) for risk of ASD associated with HOXA1/HOXB1 variants.

(A) ASD risk for HOXA1 A218G: random-effects pooled OR = 0.97, 95%CI = 0.76-1.25, P = 0.843; χ2 = 22.83, Pheterogeneity = 0.029. (B) ASD risk for HOXB1 nINS/INS: fixed-effects pooled OR = 1.14, 95%CI = 0.97-1.33, P = 0.118; χ2 = 8.78, Pheterogeneity = 0.269. Abbreviations: CC, case-control study; TDT, transmission/disequilibrium test.

https://doi.org/10.1371/journal.pone.0025603.g002

Stratified analysis

The stratified analysis was firstly performed by study design. In case-control studies, evidence of between-study heterogeneity was found for HOXA1 variant (χ2 = 13.14, Pheterogeneity = 0.069; Table 2). The pooled allelic OR in the random-effects model was 0.91 (95%CI = 0.67-1.26, P = 0.588). There was no indication of heterogeneity for HOXB1 variant (χ2 = 8.83, Pheterogeneity = 0.116; Table 3). The pooled OR of the INS allele versus nINS allele in the fixed-effects model was 1.13 (95%CI = 0.90-1.42, P = 0.284). In TDT studies, the heterogeneity test for A218G showed positive results (χ2 = 21.13, Pheterogeneity = 0.012). In the random-effect model, this variant presented no association with ASD risk (OR = 0.98, 95%CI = 0.74-1.31, P = 0.900). No sign of heterogeneity was detected for HOXB1 insertion variant (χ2 = 1.34, Pheterogeneity = 0.854), and no significant association between ASD and this variant was found in the fixed-effects model (OR = 1.16, 95%CI = 0.98-1.38, P = 0.079).

thumbnail
Table 2. Meta-analysis of case-control and TDT studies between the A218G in HOXA1 and ASD.

https://doi.org/10.1371/journal.pone.0025603.t002

thumbnail
Table 3. Meta-analysis of case-control and TDT studies between the insertion variant in HOXB1 and ASD.

https://doi.org/10.1371/journal.pone.0025603.t003

The data were further stratified by sample size into 2 subgroups, the large-sample-size subgroup (number of cases in case-control study or number of family in family-based study >80) and the small- or moderate-sample-size subgroup (number of cases in case-control study or number of family in family-based study ≤80). Heterogeneity was detected for A218G in the large-sample-size subgroup including 5 TDT studies (χ2 = 10.10, Pheterogeneity = 0.039) but not in the small-sample size subgroup including 3 case-control studies and 5 TDT studies (χ2 = 7.90, Pheterogeneity = 0.342). No significant association between A218G and ASD risk was found in the small- or large-sample size subgroup (OR = 1.22, 95%CI = 0.93-1.62 and OR = 0.83, 95%CI = 0.68-1.03, respectively). For HOXB1 insertion variant, no significant associations were observed in both the large-sample-size subgroup including 3 TDT studies (OR = 1.14, 95% CI = 0.94-1.36, Pheterogeneity = 0.733) and the small-sample size subgroup including 3 case-control studies and 2 TDT studies (OR = 1.14, 95% CI = 0.83-1.57, Pheterogeneity = 0.090).

In terms of ethnicity, the data were stratified into Whites, Blacks, Yellows, and Indians. No statistically significant finding was seen in Whites for A218G or nINS/INS variant. The pooled OR for the former was 0.92 (95% CI = 0.55-1.54, Pheterogeneity = 0.002) and the latter was 1.17 (95% CI = 0.95-1.47, Pheterogeneity = 0.719). The pooled OR could not be appraised in Blacks, Yellows, and Indians because of limited numbers of studies conducted in these populations.

Sensitivity analyses for combined studies of HOXA1 polymorphism

Given the significant between-study heterogeneity for HOXA1 A218G polymorphism, we conducted a sensitive meta-analysis to assess the effects of each individual study on the combined OR. A random-effect model was employed since heterogeneity was indicated. A series of combined OR with 95% CIs produced repeatedly after removal of each particular study consistently encompassed 1.0, suggesting the stability of the outcome that A218G was not associated with ASD risk (Table 4). Additionally, sensitivity analyses indicated that TDT study conducted in Italian families by Conciatori et al. [11] was the main origin of the heterogeneity for HOXA1 variant. The P value for Q test was not less than 0.1 after deletion of this study by Conciatori et al. (χ2 = 19.99, Pheterogeneity = 0.130), but the effect of 218G still lacked of significance (OR = 1.03, 95% CI = 0.89-1.19, P = 0.67).

thumbnail
Table 4. Sensitivity analysis of pooled OR combining family-based and case-control studies for HOXA1 A218G polymorphism.

https://doi.org/10.1371/journal.pone.0025603.t004

Cumulative meta-analyses

Cumulative meta-analyses of these two variants were also conducted via assortment of studies in chronologic order. Figure 3A and Figure 3B show the results from the cumulative meta-analyses for the HOXA1 A218G in the random-effects model and the HOXB1 nINS/INS in fixed-effects model. The effect of A218G and nINS/INS both tended to be null significant association over time. Moreover, the 95% CIs became increasingly narrower with increasing data, suggesting that the precision of the estimates was progressively enhanced by continual adding more studies.

thumbnail
Figure 3. Results of cumulative meta-analysis of association between HOXA1/HOXB1 variants and ASD risk.

(A) Cumulative forest plot for A218G using the random-effects model. (B) Cumulative forest plot for nINS/INS using the fixed-effects model.

https://doi.org/10.1371/journal.pone.0025603.g003

Publication bias

As reflected by funnel plots (Figure S1 and Figure S2) and Egger's test, there was no publication bias for A218G and nINS/INS variants (P = 0.564 and 0.591, respectively).

Discussion

This current meta-analysis included 113 cases, 184 controls, and 958 families concerning the A218G in HOXA1 and 113 cases, 184 controls, and 570 families concerning the insertion variant in HOXB1. Both the A218G and nINS/INS variants showed no significant association with risk of ASD. This current meta-analysis, to the best of our knowledge, firstly integrated the case-control and TDT studies to reflect the precision effect of HOXA1 and HOXB1 variants in ASD risk.

HOXA1 is a critical member of HOX gene family involved in the development of hindbrain [5]. Given the critical role in development of brain stem, HOXA1 was of special interest in ASD. In the HOXA1 gene, the A218G variant causing a histidine to arginine has been shown to disrupt the string of histidine repeats, which was believed to be the binding site of other proteins [9]. According to the function of this variant, studies continuously measure this variant in ASD risk, but with mixed or conflicting results. In this meta-analysis, neither the overall combined analysis nor the stratified analysis showed association between A218G and ASD with evidence of heterogeneity. To explore the source of heterogeneity, we further conducted stratified and sensitive analyses. Analysis stratified by sample size suggested that the heterogeneity was only presented in the large-sample-size subgroup, while sensitive analysis showed the heterogeneity was effectively removed after deletion of the TDT study in Italian by Conciatori et al. [11]. Interestingly, the large-sample-size subgroup comprised this TDT study by Conciatori et al., indicating this report was the main origin of the heterogeneity. Under review of this report, Conciatori et al. indicated a contradictory result to other included studies that the 218A allele, not the 218G allele, increased risk of ASD. After removing this study, the A218G variant still lacked significance in ASD. Moreover, effect estimates did not change significantly after in turn removing other studies, indicating the stability of this current result. The cumulative analysis gave further support to this current result by showing tendency of A218G toward to null-significant association after 2002 over time. In view of these, we are convinced that the result of the null-significant association between A218G variant and ASD is sound and reliable.

HOXB1, in conjunction with HOXA1, partly determined the placement of hindbrain segments in the proper location along the anterior-posterior axis during embryo development [5]. A 9-bp insertion variant, located in the amino-terminal coding region and introduces the tripeptide histidine-serine-alanine in the exon 1 of the HOXB1 gene [28], has been investigated by multiple studies in relation to ASD risk, but results were conflicting [9], [10], [13], [14], [15], [17], [27]. In this study, no significant association of this HOXB1 variant and ASD was identified in overall meta-analysis or stratified analyses by study design, sample size, and ethnicity under fixed-effect model. The cumulative analysis further reported the null-significant association tended to be stable as the number of studies and the sample size increased, conceivably suggesting no association of HOXB1 variant with ASD risk.

Integrating case-control and family based studies is an important aspect of genetic analysis, and combined estimates can provide an overall picture of the effect size attributable to genetic polymorphism. An illustrative application has been made by Gong et al. to Neuregulin 1 polymorphisms in schizophrenia [29]. In this meta-analysis, when the case-control and family studies were appraised independently, no significant association was found for the HOXA1 or HOXB1 variant, indicating the statistical consistency between these two types of studies. When the case-control studies were combined with family studies, statistical power was enhanced from enlargement of sample size, and the ORs and P values stably showed the null-significant associations. This meta-analysis integrating case-control with family-based studies therefore provides a straightforward mean to increase the ability to clarify the conflicted results of genetic studies.

Despite the clear strength of this study integrating case-control and family studies, some limitations merit serious consideration. Some heterogeneous natures of studies, including mixed population samples, variant age range of controls in case-control design, multiplex or simplex trios utilized in family studies, possibly have effect on the current result. However, due to lack of detail data or limited small number of some studies, we were unable to perform further analysis. Given the HOXB1 and HOXA1 genes have synergized in patterning hindbrain structures, there was a hypothesis that gene-gene interaction between HOXA1 and HOXB1 variants might be involved in ASD. However, interaction of these two variants could not be appraisal in this meta-analysis because of a lack of special data.

In conclusion, this meta-analysis helped for strongly clarifying the discrepancies of genetics studies into associations of HOXA1 and HOXB1 variants with ASD, and suggest null association of the A218G and nINS/INS variants with ASD risk. Further analysis should be imposed in possible interaction effect between HOXA1 and HOXB1 genetic polymorphisms in modulation of ASD risk.

Supporting Information

Table S1.

PRISMA Checklist for this meta-analysis.

https://doi.org/10.1371/journal.pone.0025603.s001

(DOC)

Figure S1.

The funnel plots of the HOXA1 A218G variant for ASD. Egger's test: P = 0.564.

https://doi.org/10.1371/journal.pone.0025603.s002

(TIF)

Figure S2.

The funnel plots of the HOXB1 nINS/INS variant for ASD. Egger's test: P = 0.591.

https://doi.org/10.1371/journal.pone.0025603.s003

(TIF)

Author Contributions

Conceived and designed the experiments: RS LZ LL XM. Performed the experiments: RS LZ RZ XZ BZ WC LL. Analyzed the data: RS LZ RZ XZ BZ WC LL. Contributed reagents/materials/analysis tools: RZ XZ BZ WC. Wrote the paper: RS LZ LL XM.

References

  1. 1. American Psychiatric Association (1994) Diagnostic and Statistical Manual of Mental Disorders. Washington, DC: American Psychiatric Press.
  2. 2. Baird G, Simonoff E, Pickles A, Chandler S, Loucas T, et al. (2006) Prevalence of disorders of the autism spectrum in a population cohort of children in South Thames: the Special Needs and Autism Project (SNAP). Lancet 368: 210–215.
  3. 3. Fombonne E (2003) The prevalence of autism. JAMA 289: 87–89.
  4. 4. Freitag CM (2007) The genetics of autistic disorders and its clinical relevance: a review of the literature. Mol Psychiatry 12: 2–22.
  5. 5. Murphy P, Hill RE (1991) Expression of the mouse labial-like homeobox-containing genes, Hox 2.9 and Hox 1.6, during segmentation of the hindbrain. Development 111: 61–74.
  6. 6. Rodier PM, Ingram JL, Tisdale B, Nelson S, Romano J (1996) Embryological origin for autism: developmental anomalies of the cranial nerve motor nuclei. J Comp Neurol 370: 247–261.
  7. 7. Rossel M, Capecchi MR (1999) Mice mutant for both Hoxa1 and Hoxb1 show extensive remodeling of the hindbrain and defects in craniofacial development. Development 126: 5027–5040.
  8. 8. Courchesne E (1997) Brainstem, cerebellar and limbic neuroanatomical abnormalities in autism. Curr Opin Neurobiol 7: 269–278.
  9. 9. Ingram JL, Stodgell CJ, Hyman SL, Figlewicz DA, Weitkamp LR, et al. (2000) Discovery of allelic variants of HOXA1 and HOXB1: genetic susceptibility to autism spectrum disorders. Teratology 62: 393–405.
  10. 10. Gallagher L, Hawi Z, Kearney G, Fitzgerald M, Gill M (2004) No association between allelic variants of HOXA1/HOXB1 and autism. Am J Med Genet B Neuropsychiatr Genet 124B: 64–67.
  11. 11. Conciatori M, Stodgell CJ, Hyman SL, O'Bara M, Militerni R, et al. (2004) Association between the HOXA1 A218G polymorphism and increased head circumference in patients with autism. Biol Psychiatry 55: 413–419.
  12. 12. Yu SY, Du YS, Wang DX, Qian YP, Jiang SD (2004) Association study between HOXA1 A218G polymorphism and autism. Journal of Chinese Medicine Research 4: 675–676.
  13. 13. Talebizadeh Z, Bittel DC, Miles JH, Takahashi N, Wang CH, et al. (2002) No association between HOXA1 and HOXB1 genes and autism spectrum disorders (ASD). J Med Genet 39: e70.
  14. 14. Sen B, Sinha S, Ahmed S, Ghosh S, Gangopadhyay PK, et al. (2007) Lack of association of HOXA1 and HOXB1 variants with autism in the Indian population. Psychiatr Genet 17: 1.
  15. 15. Romano V, Cali F, Mirisola M, Gambino G, R DA, et al. (2003) Lack of association of HOXA1 and HOXB1 mutations and autism in Sicilian (Italian) patients. Mol Psychiatry 8: 716–717.
  16. 16. Muscarella LA, Guarnieri V, Sacco R, Militerni R, Bravaccio C, et al. (2007) HOXA1 gene variants influence head growth rates in humans. Am J Med Genet B Neuropsychiatr Genet 144B: 388–390.
  17. 17. Li J, Tabor HK, Nguyen L, Gleason C, Lotspeich LJ, et al. (2002) Lack of association between HoxA1 and HoxB1 gene variants and autism in 110 multiplex families. Am J Med Genet 114: 24–30.
  18. 18. Devlin B, Bennett P, Cook EH Jr, Dawson G, Gonen D, et al. (2002) No evidence for linkage of liability to autism to HOXA1 in a sample from the CPEA network. Am J Med Genet 114: 667–672.
  19. 19. Collins JS, Schroer RJ, Bird J, Michaelis RC (2003) The HOXA1 A218G polymorphism and autism: lack of association in white and black patients from the South Carolina Autism Project. J Autism Dev Disord 33: 343–348.
  20. 20. Pabalan NA (2010) Meta-analysis in cancer genetics. Asian Pac J Cancer Prev 11: 33–38.
  21. 21. Lander ES, Schork NJ (1994) Genetic dissection of complex traits. Science 265: 2037–2048.
  22. 22. Kazeem GR, Farrall M (2005) Integrating case-control and TDT studies. Ann Hum Genet 69: 329–335.
  23. 23. Nicodemus KK (2008) Catmap: case-control and TDT meta-analysis package. BMC Bioinformatics 9: 130.
  24. 24. Moher D, Liberati A, Tetzlaff J, Altman DG (2009) Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 6: e1000097.
  25. 25. Hedges LV, Olkin I, editors. (1985) Statistical methods for meta-analysis:. Boston, London: Academic Pres Inc.
  26. 26. Egger M, Davey Smith G, Schneider M, Minder C (1997) Bias in meta-analysis detected by a simple, graphical test. BMJ 315: 629–634.
  27. 27. Muscarella LA, Guarnieri V, Sacco R, Curatolo P, Manzi B, et al. (2010) Candidate gene study of HOXB1 in autism spectrum disorder. Mol Autism 1: 9.
  28. 28. Faiella A, Zortea M, Barbaria E, Albani F, Capra V, et al. (1998) A genetic polymorphism in the human HOXB1 homeobox gene implying a 9bp tandem repeat in the amino-terminal coding region. Mutations in brief no. 200. Online. Hum Mutat 12: 363.
  29. 29. Gong YG, Wu CN, Xing QH, Zhao XZ, Zhu J, et al. (2009) A two-method meta-analysis of Neureguling1 (NRG1) association and heterogeneity in schizophrenia. Schizophrenia Research 111: 109–114.