| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Departments of Preventive Medicine (C.A.D.-D., E.D.T., V.K.C.), Urology (E.D.T.), and Obstetrics, Gynecology, and Medicine (R.Z.S.), Keck School of Medicine of the University of Southern California, Los Angeles, California 90033
Address all correspondence and requests for reprints to: Victoria Cortessis, Ph.D., Department of Preventive Medicine, Keck School of Medicine, University of Southern California, 1441 Eastlake Avenue, MC-9175, Los Angeles, California 90033. E-mail: cortessi{at}usc.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Objective: Our goals were to summarize published data on associations between AR CAG and GGC repeat lengths and male infertility and investigate sources of variation between study results.
Data Sources: We searched for reports published before October 2006 using Medline, PubMed, and Web of Science.
Study Selection: All selected studies included the following: a case group with infertility as measured by semen parameters, a control group of known or presumed fertile men, and measurement of CAG and/or GGC repeat lengths among cases and controls. Thirty-nine reports were selected based on these criteria, and 33 were ultimately included in the meta-analysis.
Data Extraction: One investigator extracted data on sample size, mean and SD of trinucleotide repeat length, and study characteristics.
Data Synthesis: Estimates of the standardized mean difference (95% confidence interval) were 0.19 (0.09–0.29) for the 33 studies and 0.31 (0.14–0.47) for a subset of 13 studies that used more stringent case and control selection criteria. Thus, in both groups, cases had statistically significantly longer CAG repeat length than controls. Publication date appeared to be a significant source of variation between studies.
Conclusions: This meta-analysis provides support for an association between increased androgen receptor CAG length and idiopathic male infertility, suggesting that even subtle disruptions in the androgen axis may compromise male fertility.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Androgens are required for male sex determination, development, and spermatogenesis. Androgen activity is mediated by the androgen receptor (AR), a member of the steroid receptor superfamily. Receptor variants with diminished capacity to respond to androgens result in androgen resistance, which compromises spermatogenesis. Additional features can also be present, with severity depending on the extent to which AR function is impaired. In the most severe form, complete androgen insensitivity syndrome, individuals with XY karyotype have female phenotype, primary amenorrhea and markedly elevated levels of serum testosterone and estradiol. In partial androgen insensitivity syndrome (Reifenstein’s syndrome), patients have ambiguous genitalia (2). In the mildest form, patients with normal male phenotype have abnormal spermatogenesis (3, 4). Based on androgen binding assays of fibroblasts from infertile men, it has been estimated that androgen resistance may be present in 40% or more of patients with idiopathic male infertility (5).
The AR is encoded by the AR gene, located on chromosome Xq11–12. The AR contains eight exons that encode three functional domains of the receptor: transactivation domain (exon 1), DNA binding domain (exons 2 and 3), and ligand-binding domain (exons 4–8) (6). Rare mutations that result in complete or partial androgen insensitivity syndromes have been localized to the ligand-binding and DNA-binding domains (4). The transactivation domain controls transcription of target genes. Two trinucleotide polymorphisms in this domain vary in length in the population: a CAG repeat encoding a polyglutamine tract and a GGC (GGN) repeat encoding a polyglycine tract.
Experimental research suggests that the number of repeats in the CAG tract is inversely correlated with transcriptional activity of the AR protein (7). The usual range in repeat length is nine to 36 repeats (8). Clinical findings have linked polyglutamine lengths of more than 40 repeats with reduced virilization and defective spermatogenesis among men affected by spinal bulbar muscular atrophy, a fatal neuromuscular disease (9). Based on this evidence, it is postulated that men with longer CAG repeats within the normal range may have subtle decreases in AR function that result in reduced spermatogenesis.
Results of studies investigating this hypothesis are widely divergent. Some report associations between infertility and longer repeats (1, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24), whereas others do not (25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46). It is unknown whether differences between these studies, including race/ethnicity of study participants and inconsistencies in case and control inclusion criteria, are responsible for conflicting findings. This possibility can be investigated in meta-analyses that include statistical measures of heterogeneity.
To our knowledge, no meta-analysis has been conducted to date analyzing results of all published studies on this association. Two prior meta-analyses (19, 32) and one pooled analysis (21) addressed subsets of published studies (12, six, and five studies, respectively) and did not quantitatively investigate the impact of heterogeneity between studies on the overall effect estimate. Our goals in preparing this report were to conduct a comprehensive meta-analysis of the published literature summarizing data on associations between AR repeat length polymorphisms and male infertility and investigate sources of heterogeneity that may have influenced published results.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We searched Medline and PubMed for articles published in English until October 2006 describing associations between male infertility and CAG and/or GGC trinucleotide repeat lengths in the AR. Search terms queried were: androgen receptor, male infertility, semen analysis, polyglutamine, polyglycine, CAG, GGC, and GGN. We screened identified publications by reviewing titles and abstracts. Bibliographies of all original reports and review articles were examined, and each was subjected to a citation search using Web of Science to identify additional publications not retrieved through online searches.
Publications identified by any of the above procedures were reviewed and then selected for possible inclusion in the meta-analysis if they fulfilled each of three criteria: 1) included a case group with infertility as measured by semen parameters based on World Health Organization guidelines (47), 2) included a control group of known or presumed fertile men, and 3) reported measurement of CAG and/or GGC repeat lengths among cases and controls. Thirty-nine reports met these criteria.
Data extraction
A single reviewer extracted data from each of the 39 reports. The following qualitative characteristics were noted: geographic location of the study population, demographic characteristics of study participants (age and race/ethnicity), case and control definitions, case and control exclusion criteria, and publication year. Quantitative data extracted were sample size and mean and SD of trinucleotide (CAG and/or GGC) repeat length for each group of cases and controls. Data were either extracted directly from articles or calculated using information provided in tables and figures. For several reports (1, 14, 16, 18, 23, 33, 36, 39, 43), SD was calculated from the SE (SD = 
*SE). One report (28)did not provide the data needed to calculate SD, so it was estimated by using the P value of the unpooled t test comparison of means between cases and controls:
 |
Data analysis
We implemented meta-analysis using Stata statistical software (Stata/SE 9.0; Stata Corp., College Station, TX). The overall standardized mean difference (SMD) and 95% confidence interval (CI) were calculated to estimate differences in repeat length between cases and controls. To determine the SMD, mean differences in number of repeats between cases and controls in each study were weighted by sample size. A random effects model was used, taking into account within-study and between-study variability. We graphically displayed the SMD along with mean differences and CIs from each study in a Forrest plot and assessed the possibility of publication bias using Egger’s unweighed regression asymmetry test (49).
To examine dispersion of data, we created Begg’s funnel plots, which display for each study the SMD vs. the SE of the SMD. Results distributed within the funnel defined by 95% confidence limits can be interpreted as variation due to sampling error. Variation due to differences in design and conduct of the studies is termed statistical heterogeneity and may result in overdispersion of results (e.g. outside the confidence limits). We used four methods to investigate potential sources of heterogeneity.
First, to learn whether the use of stricter definitions of fertility influenced results of the meta-analysis, we identified a subset of 13 studies (1, 13, 15, 16, 19, 20, 21, 23, 27, 29, 36, 38, 43) that used more stringent case and control criteria. Cases with known causes of infertility (including obstruction, infections, anatomic defects, defined genetic or endocrine disorders, and chromosomal abnormalities) were excluded from these studies; and controls were confirmed to have either sperm concentration greater than 20 x 106/ml and/or to have reported paternity of one or more children by natural conception. We further restricted cases to those with semen concentration less than 20 x 106/ml [in accordance with World Health Organization guidelines (47)], including in the subset only studies that provided this information. We did not consider sperm motility and morphology because these parameters were rarely reported. For this subset of studies, we calculated the overall SMD and 95% CI as described above and created Forrest and Begg’s funnel plots.
Second, to explore possible effects of other study characteristics, we conducted a series of analyses stratified individually on race/ethnicity of study participants [Caucasian, Asian, study population composed of several racial/ethnic groups (i.e. mixed), or unspecified]; geographic location of the study population (Europe, Asia, United States, or other); and type of control group [proven fathers and/or normozoospermic men, fertile men (no evidence of fertility specified), or unselected men]. Stratified analyses were conducted on both the full set and the subset of 13 studies. In the subset, type of control group was stratified into fathers vs. normozoospermic men.
Using data from studies that provided mean repeat length of specific case groups, we calculated SMDs to compare azoospermic cases (no sperm) and oligozoospermic cases (sperm concentration > 0 to < 20 x 106/ml) separately with controls. For each group of cases, the SMD and 95% CI were calculated.
Third, we quantified the degree of heterogeneity by calculating the I2 statistic, which estimates the proportion of variation in SMDs that is due to heterogeneity between studies, as opposed to sampling variation (50). I2 ranges from 0 to 100%, with higher values indicating greater degrees of heterogeneity (0–30%, mild heterogeneity; 30–50%, moderate heterogeneity; 50–100%, notable heterogeneity) (50). I2 was calculated from the Q-statistic, a 
2 statistic used to test for the presence of heterogeneity in meta-analyses [I2 = (Q – degrees of freedom)/Q] (50). We calculated I2 statistics for the overall analyses of the full set and the subset of 13 studies and for the stratified analyses.
Fourth, we conducted meta-regression analyses on the full set and the subset of 13 studies to investigate effects of individual study characteristics on the SMD while controlling for effects of other study characteristics. The SMD was modeled as the outcome weighted on the SE of the SMD, and study characteristics that may influence heterogeneity were included as covariates in each of two models. In model I covariates were race/ethnicity (Caucasian vs. other), geographic location (Europe vs. other), and type of control (fathers vs. all others). In model II publication date was added to the covariates in model I. We considered covariates with P < 0.05 to be modifiers of the effect of trinucleotide repeat length on the risk of infertility and therefore to be possible sources of heterogeneity.
To investigate trends in case and control repeat length over time, we conducted separate linear regression analyses of case repeat length and control repeat length on publication date.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Of the 39 articles identified, 38 reported data on the CAG repeat (1, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46), five on both the CAG and GGC repeats (10, 28, 38, 41, 46), and one on the GGC repeat (51). Among studies conducted on the GGC repeat, none reported statistically significant associations between GGC repeat length and infertility. Only two provided data required for the meta-analysis (38, 51), and data for a third was provided by the author (41). Due to the scant data available, no formal meta-analysis was conducted on the GGC repeat. Among articles addressing the CAG repeat, four were excluded because they did not provide the required data, and no additional information was received from authors (10, 25, 30, 46).
In all, data from 33 independent studies on the CAG repeat were included in this meta-analysis (1, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 27, 28, 29, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45). One article reported on two independent study groups, one from the United States and one from Singapore, so these data were included as two separate case-control series (14). Two articles compared the same control group to each of two case series (19, 31). Data reported in these articles were analyzed as follows: in most analyses, data from the larger series (19) were used; however, data from the smaller series (31) were used in the analyses of case subgroups (azoospermic and oligozoospermic) because this information was not reported for the larger series. Two additional articles (26, 28) presented data on the same case-control series, so data from only one report (28) were included. Altogether, data for 3027 cases and 2722 controls extracted from 33 reports were included in these analyses.
Characteristics of all 39 articles selected for possible inclusion are shown in Table 1
. Publication dates ranged from 1997 to 2006. Among the 33 studies included in the meta-analysis, racial/ethnic backgrounds of study participants were diverse: 17 studies enrolled Caucasian men, seven enrolled Asian men, five enrolled men of mixed races, and four did not specify race/ethnicity of the men enrolled. Study participants were enrolled in numerous geographic locations: 15 studies were conducted in Europe, seven in Asia, four in the United States, and seven in other countries. In most reports, the authors specified that cases and controls were of similar racial/ethnic background and age.
|
Analysis of the full set of 33 studies revealed statistically significantly longer CAG repeat length among cases, compared with controls (SMD = 0.19; 95% CI, 0.09–0.29) (Table 2A), as illustrated by the Forrest plot of results (Fig. 1A). The corresponding funnel plot shows overdispersion of the data (Fig. 1B), an indication of greater differences between studies than expected from sampling variation alone. Egger’s test for publication bias was significant for the full set of studies (P = 0.04).
|
|
). The corresponding Forrest plot suggests a decrease over time in the mean difference between cases and controls (Fig. 2A). The funnel plot reveals greater dispersion of the data than expected from sampling variation alone (Fig. 2B), but Egger’s test for publication bias was not statistically significant (P = 0.40).
|
Statistical assessment of heterogeneity
Stratified analyses of the full set of studies revealed differences in SMDs between some subgroups defined by race, geographic location, and control type. SMDs were slightly larger for Asian and mixed-race populations than Caucasian populations, but differences were not statistically significant (P = 0.68) (Table 2A). There were statistically significant differences between the SMDs calculated for studies conducted in Europe, Asia, the United States, and other countries (P = 0.02). Studies using proven fathers or confirmed normozoospermic men as controls found greater differences between cases and controls than studies that used other control types, but these differences were not statistically significant (P = 0.15). Among the subset of 13 studies, no significant differences in SMDs were detected between strata defined by race or geographic location (P = 0.82 and 0.76, respectively), but marginally significant differences were found for control type (P = 0.06) (Table 2B).
Specific data on azoospermic and oligozoospermic cases were provided by 20 (1, 13, 14, 15, 18, 21, 22, 27, 29, 31, 32, 33, 36, 37, 38, 39, 40, 41, 43) and 15 (1, 14, 15, 22, 24, 27, 32, 36, 37, 38, 39, 40, 41, 43) studies, respectively. Among both azoospermic and oligozoospermic cases, repeat lengths were significantly longer than among controls. However, SMD estimates for both types of cases were similar in magnitude to the overall SMD for all 33 studies (Table 2A). Results were similar when data were restricted to studies that used more stringent case and control definitions (Table 2B).
I2 statistics calculated for unstratified analyses of the full set and the subset of studies were 69 and 64%, respectively, indicating that more than half of the variation in SMDs may be due to between-study heterogeneity (Table 2). In analyses stratified on race/ethnicity, geographic location, control type, and case type, I2 statistics ranged from 14 to 88%, indicating that a notable amount of heterogeneity remained within strata.
Meta-regression analyses addressing joint effects of multiple study characteristics identified race and geographic location as significant modifiers of the SMD in all 33 studies (P = 0.001 and 0.03, respectively using model I; P = 0.02 and 0.001, respectively, using model II) but not in the subset of 13 (Table 3). Modification by publication date was highly significant in the subset of 13 studies (P = 0.005). To understand the influence of publication date in this subset better, we conducted separate linear regression analyses of case and control repeat length on publication date. There was a highly significant decrease in repeat length over time among cases (P = 0.009) but no apparent time trend among controls (P = 0.70) (data not shown).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Spermatogenesis is regulated by androgens in a largely paracrine fashion. Leydig cells of the adult testis secrete testosterone, but adult germ cells reportedly do not express the AR. Therefore, AR-mediated effects of androgens on spermatogenesis must involve the action of somatic cells. Experimental research has shown that targeted disruption of AR expression only in Sertoli cells creates mouse models with the key features of idiopathic male infertility: phenotypically normal males with severely disrupted spermatogenesis (53, 54). It is therefore reasonable to speculate that AR variants with limited Sertoli cell function may contribute to spermatogenetic deficits in men with idiopathic infertility. Moreover, because longer polyglutamine tracts appear to reduce AR function far less than mutations that cause defined androgen insensitivity syndromes, our results suggest that other determinants of subtle variation in androgen response may also influence male fertility.
This meta-analysis not only substantiates an association between CAG repeat length and infertility but also identifies sample size and differences in study design as sources of variation between earlier reports. To achieve 80% power to detect an SMD of magnitude estimated by the meta-analysis (SMD = 0.20, SD of repeat length = 3.0), 3533 cases and 3533 controls are needed (55). Although the aggregate data addressed in the meta-analysis approach this sample size, samples used in each of the 33 individual studies were extremely small by comparison.
Stringency of case and control definitions is an important determinant of differences in repeat length between cases and controls, as estimated by the SMD. Meta-analysis revealed a steady increase in the SMD as we examined data sets defined by increasingly strict definitions: among 20 studies that did not use stringent definitions, there was no statistical evidence of a difference between cases and controls. When these data were combined with those from 13 studies that used more stringent definitions, cases were found to have significantly longer CAG repeat length than controls. Even larger SMDs were observed when the subset of 13 studies was analyzed separately, particularly when controls were restricted to proven fathers (SMD = 0.37; 95% CI, 0.14–0.60). We anticipate that even this value underrepresents the difference in CAG repeat length that influences male infertility because among men with idiopathic infertility there is inevitably an unknown proportion whose infertility does not involve this polymorphism.
Stratified and meta-regression analyses identified only publication date as an additional source of variation within the subset of 13 studies, with estimated SMDs tending to increase over time (Fig. 2A). Repeat lengths among controls were nearly constant, suggesting that investigators sampled controls from similar populations over time. However, average repeat length among cases declined during the interval 1999–2005. This decline may be attributable to changing patterns of referral to infertility clinics during this period, with the introduction of new therapies such as intracytoplasmic sperm injection influencing men with a wider array of conditions to seek treatment.
To bring results of this meta-analysis to clinical decision making, answers to three questions are desired: 1) what range of AR CAG repeat lengths predisposes to idiopathic infertility; 2) what risk of infertility is associated with each length in this range; and 3) will AR-associated predisposition to infertility be transmitted to offspring conceived by in vitro fertilization using sperm of infertile men with longer repeats? The summary nature of published data included in the meta-analysis does not permit us to address questions 1 and 2 in this analysis. Therefore, collection of data required to answer these questions is now a priority. As a refinement to envisioned research, we recommend measurement of additional genotypic variants in the AR, including single nucleotide polymorphisms and the GGC repeat sequence. These data will allow investigators to address the possibility that multiple variants in the AR may act in conjunction to influence fertility and to rule out the possibility that the association reported here is substantially influenced by unmeasured variants in linkage disequilibrium with longer CAG repeats. Because the AR is located on the X chromosome, a man’s copy of the AR is normally transmitted to all of his daughters but none of his sons. Therefore, any predisposition to infertility encoded by the AR is predicted to be transmitted by a man to none of his sons and, on expectation, to one quarter of his grandsons.
In conclusion, results of this comprehensive meta-analysis suggest that variation in the AR polyglutamine tract may be a determinant of infertility in otherwise healthy men. Because longer polyglutamine tracts are far more common than mutations associated with complete or partial androgen insensitivity syndromes, this polymorphism may influence fertility in a much larger proportion of men. In light of this result, studies providing empiric estimates of the risk of infertility associated with individual tract lengths are now a pressing priority.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Disclosure Statement: The authors have no conflicts to disclose.
First Published Online August 7, 2007
Abbreviations: AR, Androgen receptor; CI, confidence interval; SMD, standardized mean difference.
Received May 18, 2007.
Accepted August 1, 2007.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
polymorphisms and androgen receptor CAG trinucleotide repeats with male infertility: a study in 109 Greek infertile men. Int J Androl 25:149–152[CrossRef][Medline]This article has been cited by other articles:
 |
|
 |
|
  K. Stouffs, H. Tournaye, I. Liebaers, and W. Lissens Male infertility and the involvement of the X chromosome Hum. Reprod. Update, June 10, 2009; (2009) dmp023v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. A Hughes and C. L Acerini Factors controlling testis descent Eur. J. Endocrinol., December 1, 2008; 159(suppl_1): S75 - S82. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. G. Martinez-Garza, M. C. Gallegos-Rivas, M. Vargas-Maciel, J. M. Rubio-Rubio, M. E. de los Monteros-Rodriguez, C. Gonzalez-Ortega, P. Cancino-Villarreal, L. G. V. de Lara, and A. M. Gutierrez-Gutierrez Genetic Screening in Infertile Mexican Men: Chromosomal Abnormalities, Y Chromosome Deletions, and Androgen Receptor CAG Repeat Length J Androl, November 1, 2008; 29(6): 654 - 660. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. A. Shah, H. J. Antoine, M. Pall, K. D. Taylor, R. Azziz, and M. O. Goodarzi Association of Androgen Receptor CAG Repeat Polymorphism and Polycystic Ovary Syndrome J. Clin. Endocrinol. Metab., May 1, 2008; 93(5): 1939 - 1945. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Lappalainen, P. Utriainen, T. Kuulasmaa, R. Voutilainen, and J. Jaaskelainen Androgen Receptor Gene CAG Repeat Polymorphism and X-Chromosome Inactivation in Children with Premature Adrenarche J. Clin. Endocrinol. Metab., April 1, 2008; 93(4): 1304 - 1309. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
