Local literature bias in genetic epidemiology: an empirical evaluation of the Chinese literature

Prediction of cardiovascular disease outcomes and established cardiovascular risk factors by genome-wide association markers

BACKGROUND

Genome-wide association (GWA) platforms have yielded a rapidly increasing number of new genetic markers. The ability of these markers to improve prediction of clinically important outcomes is debated.

METHODS AND RESULTS

A systematic review was performed of GWA-derived markers associated with cardiovascular outcomes or other phenotypes that represent common established risk factors for cardiovascular outcomes. Sources of information included the National Human Genome Research Institute catalog of published GWA studies, and perusal of the eligible GWA articles, meta-analyses on the respective associations, and articles on the incremental predictive performance of common variants in the GWA era. A total of 95 eligible associations were retrieved from the National Human Genome Research Institute catalogue of published GWA studies as of September 2008. Of those 36 have statistical support of P<10(-7). In depth evaluation of the respective articles shows 28 independent associations with such statistical support, pertaining to coronary artery disease, myocardial infarction, atrial fibrillation/flutter, prolongation of QT interval, as well as type 2 diabetes, body mass index, high-density lipoprotein levels, low-density lipoprotein levels, and nicotine dependence. Between-study heterogeneity is not taken into account usually, but it seems common and it would pose a challenge to generalizability across different populations for these markers. Still limited data are available in non-white populations. Effect sizes are small and may be even smaller in subsequent replications and meta-analysis. Population attributable fractions are substantial, given the large frequency of the risk alleles. However, individualized risk measures are typically very small (proportion of variance explained <1% per marker). When used in conjunction with traditional predictors, improvement in overall prediction (eg, area under the curve) or risk reclassification is limited, and subject to methodological caveats.

CONCLUSIONS

Despite very promising signals in terms of statistical significance, evidence for improvement in cardiovascular prediction by currently available markers derived from GWA studies is sparse. Clinical use of such markers currently would be premature.

Evaluation of the potential excess of statistically significant findings in published genetic association studies: application to Alzheimer's disease

The authors evaluated whether there is an excess of statistically significant results in studies of genetic associations with Alzheimer's disease reflecting either between-study heterogeneity or bias. Among published articles on genetic associations entered into the comprehensive AlzGene database (www.alzgene.org) through January 31, 2007, 1,348 studies included in 175 meta-analyses with 3 or more studies each were analyzed. The number of observed studies (O) with statistically significant results (P = 0.05 threshold) was compared with the expected number (E) under different assumptions for the magnitude of the effect size. In the main analysis, the plausible effect size of each association was the summary effect presented in the respective meta-analysis. Overall, 19 meta-analyses (all with eventually nonsignificant summary effects) had a documented excess of O over E: Typically single studies had significant effects pointing in opposite directions and early summary effects were dissipated over time. Across the whole domain, O was 235 (17.4%), while E was 164.8 (12.2%) (P < 10(-6)). The excess showed a predilection for meta-analyses with nonsignificant summary effects and between-study heterogeneity. The excess was seen for all levels of statistical significance and also for studies with borderline P values (P = 0.05-0.10). The excess of significant findings may represent significance-chasing biases in a setting of massive testing.

Why current publication practices may distort science

John Ioannidis and colleagues argue that the current system of publication in biomedical research provides a distorted view of the reality of scientific data.

Calibration of credibility of agnostic genome-wide associations

Genome-wide testing platforms are increasingly used to promote "agnostic" approaches to the discovery of gene variants associated with the risk of many common diseases and quantitative traits. The early track record of genome-wide association (GWA) studies suggests that some proposed associations are replicated quite consistently with large-scale subsequent evidence from multiple studies, others have a more inconsistent replication record, some have failed to be replicated by independent investigators and many more early proposed associations await further replication. An important question is how to calibrate the credibility of these postulated associations. A simple Bayesian method is applied here to achieve such calibration. The variability of the estimated credibility is examined under different assumptions. Empirical examples are drawn from existing GWA studies. It is demonstrated that the credibility of different proposed associations can cover a very wide range. The credibility of specific associations usually remains relatively robust when different plausible assumptions are made (within a reasonable range) for the prior odds of an association being true, or the magnitude of the anticipated effect size for genetic associations. Heterogeneity and bias assumptions can have a more major impact on the credibility estimates and thus they need very careful consideration in each case. Credibility calibration may be used in conjunction with qualitative criteria for the appraisal of the cumulative evidence that take into consideration the amount, consistency, and protection from bias in the data.

Why most discovered true associations are inflated

Newly discovered true (non-null) associations often have inflated effects compared with the true effect sizes. I discuss here the main reasons for this inflation. First, theoretical considerations prove that when true discovery is claimed based on crossing a threshold of statistical significance and the discovery study is underpowered, the observed effects are expected to be inflated. This has been demonstrated in various fields ranging from early stopped clinical trials to genome-wide associations. Second, flexible analyses coupled with selective reporting may inflate the published discovered effects. The vibration ratio (the ratio of the largest vs. smallest effect on the same association approached with different analytic choices) can be very large. Third, effects may be inflated at the stage of interpretation due to diverse conflicts of interest. Discovered effects are not always inflated, and under some circumstances may be deflated-for example, in the setting of late discovery of associations in sequentially accumulated overpowered evidence, in some types of misclassification from measurement error, and in conflicts causing reverse biases. Finally, I discuss potential approaches to this problem. These include being cautious about newly discovered effect sizes, considering some rational down-adjustment, using analytical methods that correct for the anticipated inflation, ignoring the magnitude of the effect (if not necessary), conducting large studies in the discovery phase, using strict protocols for analyses, pursuing complete and transparent reporting of all results, placing emphasis on replication, and being fair with interpretation of results.

Pharmacogenetics of the response to beta 2 agonist drugs: a systematic overview of the field

The response to beta2-agonist treatment shows large repeatability within individuals and may thus be determined by genetic influences. Here we present a systematic overview of the available genetic association and linkage data for beta2-agonist treatment response. Systematic searches identified 66 eligible articles, as of March 2007, pertaining either to B2AR gene polymorphisms and short-acting or long-acting beta2-agonists or to another 29 different genes. We systematize these study results according to gene, agent and type of outcomes addressed. The systematic review highlights major challenges in the field, including extreme multiplicity of analyses; lack of consensus for main phenotypes of interest; typically small sample sizes; and poor replicability of the proposed genetic variants. Future studies will benefit from standardization of analyses and outcomes, hypothesis-free genome-wide association testing platforms, potentially additional fine mapping around new discovered variants, and large-scale collaborative studies with prospective plans for replication among several teams, with transparent public recording of all data.

Required sample size and nonreplicability thresholds for heterogeneous genetic associations

Many gene-disease associations proposed to date have not been consistently replicated across different populations. Nonreplication often reflects false positives in the original claims. However, occasionally, nonreplication may be due to heterogeneity due to biases or even genuine diversity of the genetic effects in different populations. Here, we propose methods for estimating the required sample size to replicate an association across many studies with different amounts of between-study heterogeneity, when data are summarized through metaanalysis. We demonstrate thresholds of between-study heterogeneity (tau(0)(2)) above which one cannot reach adequate power to replicate a proposed association at a specified level of statistical significance when k studies are performed (regardless of how large these studies are). Based on empirical evidence from 91 proposed gene-disease associations (50 on candidate genes and 41 from genome-wide association efforts), the observed between-study heterogeneity is often close to or even surpasses nonreplicability thresholds. With more modest between-study heterogeneity, the required sample size increases considerably compared with when no between-study heterogeneity exists. Increases are steep as tau(0)(2) is approached. Therefore, some true associations may not be practically possible to replicate with consistency, no matter how large studies are conducted. Efforts should be made to minimize between-study heterogeneity in targeted genetic effects.

Methods for meta-analysis in genetic association studies: a review of their potential and pitfalls

Meta-analysis offers the opportunity to combine evidence from retrospectively accumulated or prospectively generated data. Meta-analyses may provide summary estimates and can help in detecting and addressing potential inconsistency between the combined datasets. Application of meta-analysis in genetic associations presents considerable potential and several pitfalls. In this review, we present basic principles of meta-analytic methods, adapted for human genome epidemiology. We describe issues that arise in the retrospective or the prospective collection of relevant data through various sources, common traps to consider in the appraisal of evidence and potential biases that may interfere. We describe the relative merits and caveats for common methods used to trace inconsistency across studies along with possible reasons for non-replication of proposed associations. Different statistical models may be employed to combine data and some common misconceptions may arise in the process. Several meta-analysis diagnostics are often applied or misapplied in the literature, and we comment on their use and limitations. An alternative to overcome limitations arising from retrospective combination of data from published studies is to create networks of research teams working in the same field and perform collaborative meta-analyses of individual participant data, ideally on a prospective basis. We discuss the advantages and the challenges inherent in such collaborative approaches. Meta-analysis can be a useful tool in dissecting the genetics of complex diseases and traits, provided its methods are properly applied and interpreted.

Assessment of cumulative evidence on genetic associations: interim guidelines

Established guidelines for causal inference in epidemiological studies may be inappropriate for genetic associations. A consensus process was used to develop guidance criteria for assessing cumulative epidemiologic evidence in genetic associations. A proposed semi-quantitative index assigns three levels for the amount of evidence, extent of replication, and protection from bias, and also generates a composite assessment of 'strong', 'moderate' or 'weak' epidemiological credibility. In addition, we discuss how additional input and guidance can be derived from biological data. Future empirical research and consensus development are needed to develop an integrated model for combining epidemiological and biological evidence in the rapidly evolving field of investigation of genetic factors.

The CASP8 -652 6N del promoter polymorphism and breast cancer risk: a multicenter study

A recent study on an Asian population reported a six-nucleotide insertion-deletion polymorphism (-652 6N del) in the CASP8 promoter region to be strongly associated with a decreased risk of multiple types of cancer, including breast cancer (BC). Here, we investigate the effect of this deletion in four independent large European BC case-control studies, including data from a total of 7,753 cases and 7,921 controls. The combined per allele odds ratio (OR) was 0.97 (95% confidence interval (CI), 95% CI = 0.93-1.02). The present result indicates that the CASP8 -652 6N del variant has no significant effect on BC risk in Europeans.

Three endothelial nitric oxide (NOS3) gene polymorphisms in hypertensive and normotensive individuals: meta-analysis of 53 studies reveals evidence of publication bias

BACKGROUND

Studies on the relationship between endothelial nitric oxide (NOS3) gene variants and hypertension have been conflicting. To explore this hypothesis further, we performed a meta-analysis and re-evaluated the relationship between the three most widely studied NOS3 polymorphisms and hypertension status and blood pressure levels.

METHODS

Data on 40,413 subjects from 53 studies were combined in five distinct meta-analyses, and heterogeneity and publication bias were explored.

RESULTS

Heterogeneity was observed in all meta-analyses. By a random-effects model, carriers of the four 27-basepair repeat variable number of tandem repeats in intron 4 were associated with a 28% increase in the risk of hypertension compared with those homozygous for the 5 repeat: odds ratio (OR) 1.28, 95% confidence interval (CI) 1.11-1.47, P=0.001. In Asian individuals, Asp allele carriers displayed a similar association: OR 1.28, 95% CI 1.06-1.54, P=0.01, as well as a 2 mmHg increase in both systolic (P=0.04) and diastolic (P=0.009) blood pressure levels. Furthermore, meta-regression analysis indicated that the effect of the Glu298Asp genotype on the risk of hypertension might be dependent on total cholesterol status. No effect of the T-786C variant on hypertension was detected. There was evidence that such findings might be a result of selectively reporting/publishing positive reports.

CONCLUSION

Our results suggest that current data on the relationship between NOS3 variants and hypertension are subject not only to important heterogeneity but also to publication bias. Future research should preferentially focus on gene-environment interactions as well as haplotype analyses.

Interleukin-1B polymorphisms and gastric cancer risk--a meta-analysis

Some studies have reported that proinflammatory polymorphisms in interleukin-1B (IL-1B) and IL-1 receptor antagonist (IL-1RN) genes are associated with increased gastric cancer risk. However, other studies have shown null or inverse associations. This meta-analysis reviews and summarizes published evidence for these associations. Searching the PubMed Database yielded 35 studies that reported on the association between IL-1B -511 C>T, IL-1B -31 T>C, or IL-1RN variable number tandem repeat polymorphisms and gastric cancer risk. Q-statistics and I(2) statistics were calculated to examine heterogeneity. Summary odds ratios (OR) and 95% confidence intervals (95% CI) were calculated in the random-effects model using the DerSimonian-Laird method. For all gastric cancers, the overall ORs (95% CIs) for IL-1B -511 CT versus CC and TT versus CC genotypes were 1.07 (0.91-1.25) and 1.16 (0.95-1.42), respectively. ORs (95% CIs) for the association between IL-1B -31 CT versus TT and CC versus TT genotypes were 0.99 (0.83-1.19) and 0.98 (0.78-1.21), respectively. For the associations between IL-1RN and gastric cancer, ORs (95% CIs) for *2/L versus LL and *2/*2 versus L/L were 1.15 (0.96-1.38) and 1.23 (0.79-1.92). For each of the examined associations, there was significant heterogeneity among studies; P(heterogeneity) < or = 0.001 and I(2) ranged from 0.54 to 0.71. Noncardia cancers showed stronger associations with IL-1B -511 CT or TT and IL1-RN *2/*2 genotypes, but limiting the analysis to intestinal-type cancers, studies conducted in Western countries, or studies in which polymorphisms were in Hardy-Weinberg equilibrium, made no material difference in the results. The overall associations between IL-1B or IL-1RN proinflammatory polymorphisms and gastric cancer were null but several studies showed an association. The sources of this variation are unclear.

No evidence for association of CTLA-4 gene polymorphisms with the risk of developing multiple sclerosis: a meta-analysis

We conducted a meta-analysis concerning the association of cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) gene polymorphisms with the risk of developing multiple sclerosis (MS). We identified 18 eligible studies summarizing information about 3375 MS cases and 2930 healthy controls. Two polymorphisms were of interest: the exon 1 +49 A/G polymorphism (in 18 studies) and the promoter-318 C/T polymorphism (in 10 studies). Using random-effects methods we found no evidence for association of the various contrasts of genotypes (or allele frequencies) with the disease. There was significant between-studies heterogeneity that could not be explained by the ethnicity of the populations studied or by other summary measures (gender, disease course, latitude). The major finding of the meta-analysis, apart from the lack of an overall association, consists of detecting a significant time trend of the OR for the contrast of GA versus GG+AA genotypes of the exon 1 +49 A/G polymorphism. In particular, using cumulative meta-analysis we found that the large number of conflicting results on the subject was triggered by the early appearance of a highly significant published result (a study that indicated a significant association of the genotype with the disease).

Non-Replication and Inconsistency in the Genome-Wide Association Setting

Non-replication and inconsistency had been common features in the search for common variants of candidate genes affecting the risk of complex diseases. They may continue to require attention in the current era, when massive hypothesis-free testing of genetic variants is feasible. An empirical evaluation of the early experience with genome-wide association (GWA) studies suggests several examples where proposed associations have failed to be replicated by subsequent investigations. Non-replication and inconsistency is defined here in the framework of cumulative meta-analysis. Ideally, associations exist, GWA finds them, and subsequent investigations should replicate them. However, a number of other possibilities need to be considered. No common genetic variants may associate with the phenotype of interest and GWA may find nothing; or associations may exist, but GWA may miss them. Associations that do not exist may be falsely selected by the GWA and subsequent studies may appropriately refute them or falsely replicate them. Finally, GWA may find true associations that are nevertheless falsely non-replicated in the subsequent studies; or associations may be genuinely inconsistent across study populations. A list of options is presented for consideration in each of these scenarios.

Genetics of ischaemic stroke among persons of non-European descent: a meta-analysis of eight genes involving approximately 32,500 individuals

BACKGROUND

Ischaemic stroke in persons of European descent has a genetic basis, but whether the stroke-susceptibility alleles, the strength of any association, and the extent of their attributable risks are the same in persons of non-European descent remains unanswered. Whether ethnicity itself has a relevant or substantial contribution on those effect estimates is controversial. Comparative analyses between the ethnic groups may allow general conclusions to be drawn about polygenic disorders.

METHODS AND FINDINGS

We performed a literature-based systematic review of genetic association studies in stroke in persons of non-European descent. Odds ratios (ORs) and 95% confidence intervals (CIs) were determined for each gene-disease association using fixed and random effect models. We further performed a comparative genetic analysis across the different ethnic groups (including persons of European descent derived from our previous meta-analysis) to determine if genetic risks varied by ethnicity. Following a review of 500 manuscripts, eight candidate gene variants were analysed among 32,431 individuals (12,883 cases and 19,548 controls), comprising mainly Chinese, Japanese, and Korean individuals. Of the eight candidate genes studied, three were associated with ischaemic stroke: the angiotensin I converting enzyme (ACE) insertion/deletion (I/D) polymorphism with a mean OR of 1.90 (95% CI 1.23-2.93) in the Chinese and 1.74 (95% CI 0.88-3.42) in the Japanese; the summary OR for the C677T variant of 5,10-methylenetetrahydrofolate reductase (MTHFR) was 1.18 (95% CI 0.90-1.56) in Chinese and 1.34 (95% CI 0.87-2.06) in Koreans; and the pooled OR for the apolipoprotein E (APOE) gene was 2.18 (95% CI 1.52-3.13) in Chinese and 1.51 (95% CI 0.93-2.45) in Japanese. Comparing the commonly investigated stroke genes among the Asian groups against studies in persons of European descent, we found an absence of any substantial qualitative or quantitative interaction for ORs by ethnicity. However, the number of individuals recruited per study in the studies of persons of non-European descent was significantly smaller compared to studies of persons of European descent, despite a similar number of studies conducted per gene.

CONCLUSIONS

These data suggest that genetic associations studied to date for ischaemic stroke among persons of non-European descent are similar to those for persons of European descent. Claims of differences in genetic effects among different ethnic populations for complex disorders such as stroke may be overstated. However, due to the limited number of gene variants evaluated, the relatively smaller number of individuals included in the meta-analyses of persons of non-European descent in stroke, and the possibility of publication bias, the existence of allele variants with differential effects by ethnicity cannot be excluded.

The GNB3 C825T polymorphism and essential hypertension: a meta-analysis of 34 studies including 14,094 cases and 17,760 controls

OBJECTIVES

The C825T single nucleotide polymorphism of the G-protein beta3 (GNB3) has been implicated in susceptibility to essential hypertension, through the expression of an alternatively spliced truncated variant. In an effort to clarify earlier inconclusive results, we performed a meta-analysis of population-based case-control genetic association studies.

METHODS

Random-effects methods were applied on summary data in order to combine the results of the individual studies.

RESULTS

We identified in total 34 studies, including 14,094 hypertensive cases and 17,760 controls. The TT versus CC + CT contrast yielded an overall odds ratio (OR) of 1.08 [95% confidence interval (CI): 1.01, 1.15], the contrast of TT + CT versus CC, an OR of 1.17 (95% CI: 1.06, 1.29), whereas that of the T allele versus C allele yielded a non-significant OR of 1.05 (95% CI: 0.98, 1.13). There was moderate evidence for a publication bias in the latter two contrasts, which was eliminated after excluding studies not in Hardy-Weinberg equilibrium and those performed on non-normal populations (those with a diagnosis of diabetes, obesity and myocardial infarction). Subgroup analyses revealed that non-significant estimates arose from studies on Asian populations, as opposed to the Caucasian ones. Furthermore, the frequency of the T allele was lower in Caucasians and these populations were found to inhabit higher latitudes.

CONCLUSIONS

The findings are in agreement with a recently proposed causal model for systolic blood pressure, which correlates it with the T allele and the absolute latitude. Further studies are needed in order to fully address questions about the aetiological mechanism of the particular association, as well as to study the effect in populations of African descent.

A critical appraisal of epidemiological studies comes from basic knowledge: a reader's guide to assess potential for biases

Scientific literature may be biased because of the internal validity of studies being compromised by different forms of measurement error, and/or because of the selective reporting of positive and 'statistically significant' results. While the first source of bias might be prevented, and in some cases corrected to a degree, the second represents a pervasive problem afflicting the medical literature; a situation that can only be 'corrected' by a change in the mindset of authors, reviewers, and editors. This review focuses on the concepts of confounding, selection bias and information bias, utilising explanatory examples and simple rules to recognise and, when possible, to correct for them. Confounding is a mixing of effects resulting from an imbalance of some of the causes of disease across the compared groups. It can be prevented by randomization and restriction, and controlled by stratification, standardization or by using multivariable techniques. Selection bias stems from an absence of comparability among the groups being studied, while information bias arises from distorted information collection techniques. Publication bias of medical research results can invalidate evidence-based medicine, when a researcher attempting to collect all the published studies on a specific topic actually gathers only a proportion of them, usually the ones reporting 'positive' results. The selective publication of 'statistically significant' results represents a problem that researchers and readers have to be aware of in order to face the entire body of published medical evidence with a degree of scepticism.

Selection in reported epidemiological risks: an empirical assessment

BACKGROUND

Epidemiological studies may be subject to selective reporting, but empirical evidence thereof is limited. We empirically evaluated the extent of selection of significant results and large effect sizes in a large sample of recent articles.

METHODS AND FINDINGS

We evaluated 389 articles of epidemiological studies that reported, in their respective abstracts, at least one relative risk for a continuous risk factor in contrasts based on median, tertile, quartile, or quintile categorizations. We examined the proportion and correlates of reporting statistically significant and nonsignificant results in the abstract and whether the magnitude of the relative risks presented (coined to be consistently > or =1.00) differs depending on the type of contrast used for the risk factor. In 342 articles (87.9%), > or =1 statistically significant relative risk was reported in the abstract, while only 169 articles (43.4%) reported > or =1 statistically nonsignificant relative risk in the abstract. Reporting of statistically significant results was more common with structured abstracts, and was less common in US-based studies and in cancer outcomes. Among 50 randomly selected articles in which the full text was examined, a median of nine (interquartile range 5-16) statistically significant and six (interquartile range 3-16) statistically nonsignificant relative risks were presented (p = 0.25). Paradoxically, the smallest presented relative risks were based on the contrasts of extreme quintiles; on average, the relative risk magnitude was 1.41-, 1.42-, and 1.36-fold larger in contrasts of extreme quartiles, extreme tertiles, and above-versus-below median values, respectively (p < 0.001).

CONCLUSIONS

Published epidemiological investigations almost universally highlight significant associations between risk factors and outcomes. For continuous risk factors, investigators selectively present contrasts between more extreme groups, when relative risks are inherently lower.

Is Molecular Profiling Ready for Use in Clinical Decision Making?

Molecular profiling, the classification of tissue or other specimens for diagnostic, prognostic, and predictive purposes based on multiple gene expression, is a technology that holds major promise for optimizing the management of patients with cancer. However, the use of these tests for clinical decision making presents many challenges to overcome. Assay development and data analysis in this field have been largely exploratory, and leave numerous possibilities for the introduction of bias. Standardization of profiles remains the exception. Classifier performance is usually overinterpreted by presenting the results as p-values or multiplicative effects (e.g., relative risks), while the absolute sensitivity and specificity of classification remain modest at best, especially when tested in large validation samples. Validation has often been done with suboptimal attention to methodology and protection from bias. The postulated classifier performance may be inflated compared to what these profiles can achieve. With the exception of breast cancer, we have little evidence about the incremental discrimination that molecular profiles can provide versus classic risk factors alone. Clinical trials have started to evaluate the utility of using molecular profiles for breast cancer management. Until we obtain data from these trials, the impact of these tests and the net benefit under real-life settings remain unknown. Optimal incorporation into clinical practice is not straightforward. Finally, cost-effectiveness is difficult to appreciate until these other challenges are addressed. Overall, molecular profiling is a fascinating and promising technology, but its incorporation into clinical decision making requires careful planning and robust evidence.

Limitations are not properly acknowledged in the scientific literature

Limitations are important to understand for placing research findings in context, interpreting the validity of the scientific work, and ascribing a credibility level to the conclusions of published research. This goes beyond listing the magnitude and direction of random and systematic errors and validity problems. Acknowledgment of limitations requires an interpretation of the meaning and influence of errors and validity problems on the published findings. An examination of the full-text files of the first 50 articles published in 2005 in the six most-cited research journals and in two recently launched leading open-access journals showed that only 67 articles (17%) used at least one word denoting limitations in the context of the presented scientific work. Only four articles (1%) used the word limitation in their abstract; none referred to limitations of the present work that materially affected conclusions. Only five articles had a separate section on limitations. Conversely, 243 articles (61%) used words detected by the roots error, valid, bias, reproducib, or false and 289 articles (72%) used words with the root importan. Among the 25 top-cited journals' instructions to the authors and editorial policies, only one encourages discussion of limitations; importance, novelty, and lack of error are typically encouraged. Limitations should be better covered and discussed in research articles. To facilitate this, journals should give better guidance and promote the discussion of limitations. Otherwise, we are facing an important loss of context for the scientific literature.

Angiotensin-converting enzyme I/D polymorphism and preeclampsia risk: evidence of small-study bias

BACKGROUND

Inappropriate activation of the renin-angiotensin system may play a part in the development of preeclampsia. An insertion/deletion polymorphism within the angiotensin-I converting enzyme gene (ACE-I/D) has shown to be reliably associated with differences in angiotensin-converting enzyme (ACE) activity. However, previous studies of the ACE-I/D variant and preeclampsia have been individually underpowered to detect plausible genotypic risks.

METHODS AND FINDINGS

A prospective case-control study was conducted in 1,711 unrelated young pregnant women (665 preeclamptic and 1,046 healthy pregnant controls) recruited from five Colombian cities. Maternal blood was obtained to genotype for the ACE-I/D polymorphism. Crude and adjusted odds ratio (OR) and 95% confidence interval (CI) using logistic regression models were obtained to evaluate the strength of the association between ACE-I/D variant and preeclampsia risk. A meta-analysis was then undertaken of all published studies to February 2006 evaluating the ACE-I/D variant in preeclampsia. An additive model (per-D-allele) revealed a null association between the ACE-I/D variant and preeclampsia risk (crude OR = 0.95 [95% CI, 0.81-1.10]) in the new case-control study. Similar results were obtained after adjusting for confounders (adjusted per-allele OR = 0.90 [95% CI, 0.77-1.06]) and using other genetic models of inheritance. A meta-analysis (2,596 cases and 3,828 controls from 22 studies) showed a per-allele OR of 1.26 (95% CI, 1.07-1.49). An analysis stratified by study size showed an attenuated OR toward the null as study size increased.

CONCLUSIONS

It is highly likely that the observed small nominal increase in risk of preeclampsia associated with the ACE D-allele is due to small-study bias, similar to that observed in cardiovascular disease. Reliable assessment of the origins of preeclampsia using a genetic approach may require the establishment of a collaborating consortium to generate a dataset of adequate size.

NAD(P)H:quinone oxidoreductase 1 (NQO1) Pro187Ser polymorphism and the risk of lung, bladder, and colorectal cancers: a meta-analysis

NAD(P)H:quinone oxidoreductase 1 (NQO1) is a cytosolic enzyme that catalyzes the two-electron reduction of quinoid compounds into hydroquinones, their less toxic form. A sequence variant at position 609 (C --> T) in the NQO1 gene encodes an enzyme with reduced quinone reductase activity in vitro and thus was hypothesized to affect cancer susceptibility. We conducted meta-analyses focusing on three cancer sites (lung, bladder, and colorectum) to summarize the findings from the current literature and to explore sources of heterogeneity.

RESULTS

There is no clear association between the NQO1 Pro187Ser polymorphism and lung cancer risk in the three ethnic groups examined: odds ratio (OR(White)) C/T + T/T versus C/C = 1.04 [95% confidence interval (95% CI), 0.96-1.13], OR(Asian) = 0.99 (95% CI, 0.72-1.34), and OR(Blacks) = 0.95 (95% CI, 0.66-1.36). However, a modestly increased risk was suggested for the variant homozygotes in whites (OR T/T versus C/C, 1.19; 95% CI, 0.94-1.50). Analysis excluding one outlier study suggested the variant allele may be associated with reduced lung cancer risk in Asians. Meta-analyses for bladder and colorectal cancer suggested a statistically significant association with the variant genotypes in whites. In stratified analyses, the NQO1 Pro187Ser variant genotypes were associated with slightly increased lung cancer risk in white ever smokers but not in white never smokers and were mainly associated with a reduced risk of lung adenocarcinoma but not squamous cell carcinoma in Asians.

CONCLUSIONS

Results from our meta-analyses suggest that the variant NQO1 Pro187Ser genotype may affect individual susceptibility to lung, bladder, and colorectal cancer. Such effects of the NQO1 polymorphism seem to be modified by ethnicity and smoking status.

Massage intervention for promoting mental and physical health in infants aged under six months

BACKGROUND

Infant massage is increasingly being used in the community for low-risk babies and their primary care givers. Anecdotal claims suggest benefits for sleep, respiration, elimination and the reduction of colic and wind. Infant massage is also thought to reduce infant stress and promote positive parent-infant interaction.

OBJECTIVES

The aim of this review was to assess the effectiveness of infant massage in promoting infant physical and mental health in population samples.

SEARCH STRATEGY

Searches were undertaken of CENTRAL 2005 (Issue 3), MEDLINE (1970 to 2005), PsycINFO (1970 to 2005), CINAHL (1982 to 2005), EMBASE (1980 to 2005), and a number of other Western and Chinese databases.

SELECTION CRITERIA

Studies in which babies under the age of six months were randomised to an infant massage or a no-treatment control group, and utilising a standardised outcome measuring infant mental or physical development.

DATA COLLECTION AND ANALYSIS

Weighted and standardised mean differences and 95% confidence intervals are presented. Where appropriate the results have been combined in a meta-analysis using a random effects model.

MAIN RESULTS

Twenty-three studies were included in the review. One was a follow-up study and thirteen were included in a separate analysis due to concerns about the uniformly significant results and the lack of dropout. The results of nine studies providing primary data suggest that infant massage has no effect on growth, but provides some evidence suggestive of improved mother-infant interaction, sleep and relaxation, reduced crying and a beneficial impact on a number of hormones controlling stress. Results showing a significant impact on number of illnesses and clinic visits were limited to a study of Korean orphanage infants. There was no evidence of effects on cognitive and behavioural outcomes, infant attachment or temperament. The data from the 13 studies regarded to be at high risk of bias show uniformly significant benefits on growth, sleep, crying and bilirubin levels.

AUTHORS' CONCLUSIONS

The only evidence of a significant impact of massage on growth was obtained from a group of studies regarded to be at high risk of bias. There was, however, some evidence of benefits on mother-infant interaction, sleeping and crying, and on hormones influencing stress levels. In the absence of evidence of harm, these findings may be sufficient to support the use of infant massage in the community, particularly in contexts where infant stimulation is poor. Further research is needed, however, before it will be possible to recommend universal provision.

Lack of association between the platelet glycoprotein Ia C807T gene polymorphism and coronary artery disease: a meta-analysis

BACKGROUND

The platelet-collagen receptor, glycoprotein (GP) la/lla plays a crucial role in the adhesion of platelets to fibrillar collagen, an event contributing to the pathogenesis of thrombosis. The C807T polymorphism of the GPla gene is considered a genetic marker of the platelet GPla/lla density. The importance of the GPla gene C807T polymorphism as a genetic risk factor for coronary artery disease (CAD) remains controversial. To assess this association, we performed a meta-analysis of published data.

METHODS

A comprehensive meta-analysis of 19 studies, with a total sample of 13835 subjects using random effects models.

RESULTS

The C versus T allele contrast gave an OR of 0.998 with 95% Cl 0.937-1.064. Similarly, comparing the C homozygotes with the T homozygotes, the CC genotype versus the others and the TT genotype versus the rest, no evidence of any gene-disease association was obtained. Furthermore, the meta-regression analysis did not disclose any variable that could modify the role of this polymorphism in the development of CAD.

CONCLUSION

Our findings support the view that C807T polymorphism of the GPla gene is not a significant risk factor for CAD, either alone or in combination with other major cardiovascular risk factors.

Implications of small effect sizes of individual genetic variants on the design and interpretation of genetic association studies of complex diseases

Accumulated evidence from searching for candidate gene-disease associations of complex diseases can offer some insights as the field moves toward discovery-oriented approaches with massive genome-wide testing. Meta-analyses of 50 non-human lymphocyte antigen gene-disease associations with documented overall statistical significance (752 studies) show summary odds ratios with a median of 1.43 (interquartile range, 1.28-1.65). Many different biases may operate in this field, for both single studies and meta-analyses, and these biases could invalidate some of these seemingly "validated" associations. Studies with a sample size of >500 show a median odds ratio of only 1.15. The median sample size required to detect the observed summary effects in each population addressed in the 752 studies is estimated to be 3,535 (interquartile range, 1,936-9,119 for cases and controls combined). These estimates are steeply inflated in the presence of modest bias. Population heterogeneity, as well as gene-gene and gene-environment interactions, could steeply increase these estimates and may be difficult to address even by very large biobanks and observational cohorts. The one visible solution is for a large number of teams to join forces on the same research platforms. These collaborative studies ideally should be designed up front to also assess more complex gene-gene and gene-environment interactions.

Family-based versus unrelated case-control designs for genetic associations

The most simple and commonly used approach for genetic associations is the case-control study design of unrelated people. This design is susceptible to population stratification. This problem is obviated in family-based studies, but it is usually difficult to accumulate large enough samples of well-characterized families. We addressed empirically whether the two designs give similar estimates of association in 93 investigations where both unrelated case-control and family-based designs had been employed. Estimated odds ratios differed beyond chance between the two designs in only four instances (4%). The summary relative odds ratio (ROR) (the ratio of odds ratios obtained from unrelated case-control and family-based studies) was close to unity (0.96 [95% confidence interval, 0.91–1.01]). There was no heterogeneity in the ROR across studies (amount of heterogeneity beyond chance I² = 0%). Differences on whether results were nominally statistically significant (p < 0.05) or not with the two designs were common (opposite classification rates 14% and 17%); this reflected largely differences in power. Conclusions were largely similar in diverse subgroup analyses. Unrelated case-control and family-based designs give overall similar estimates of association. We cannot rule out rare large biases or common small biases.

Different types of designs are used for the assessment of genetic associations for complex diseases. Case-control studies of unrelated people and family-based designs are the most widely used. Each has its advantages and disadvantages. This paper compares the estimates of the two types of design using a meta-analytic approach, i.e. a systematic selection of data and quantitative synthesis of results across many studies. The authors examined 93 associations where both unrelated case-control and family-based designs had been employed. Both designs gave overall similar estimates of association and the conclusions were very similar in subgroup analyses that considered various design features that might affect in theory the degree of agreement between the two designs. No heterogeneity between studies was observed. Hence, there was no consistent pattern of over-estimation or under-estimation of the probed association with one or the other design. However, one cannot exclude the possibility that rare large differences or common small differences may occur between the two designs.

Large-scale evidence for the effect of the COLIA1 Sp1 polymorphism on osteoporosis outcomes: the GENOMOS study

BACKGROUND

Osteoporosis and fracture risk are considered to be under genetic control. Extensive work is being performed to identify the exact genetic variants that determine this risk. Previous work has suggested that a G/T polymorphism affecting an Sp1 binding site in the COLIA1 gene is a genetic marker for low bone mineral density (BMD) and osteoporotic fracture, but there have been no very-large-scale studies of COLIA1 alleles in relation to these phenotypes.

METHODS AND FINDINGS

Here we evaluated the role of COLIA1 Sp1 alleles as a predictor of BMD and fracture in a multicenter study involving 20,786 individuals from several European countries. At the femoral neck, the average (95% confidence interval [CI]) BMD values were 25 mg/cm2 (CI, 16 to 34 mg/cm2) lower in TT homozygotes than the other genotype groups (p < 0.001), and a similar difference was observed at the lumbar spine; 21 mg/cm2 (CI, 1 to 42 mg/cm2), (p = 0.039). These associations were unaltered after adjustment for potential confounding factors. There was no association with fracture overall (odds ratio [OR] = 1.01 [CI, 0.95 to 1.08]) in either unadjusted or adjusted analyses, but there was a non-significant trend for association with vertebral fracture and a nominally significant association with incident vertebral fractures in females (OR = 1.33 [CI, 1.00 to 1.77]) that was independent of BMD, and unaltered in adjusted analyses.

CONCLUSIONS

Allowing for the inevitable heterogeneity between participating teams, this study-which to our knowledge is the largest ever performed in the field of osteoporosis genetics for a single gene-demonstrates that the COLIA1 Sp1 polymorphism is associated with reduced BMD and could predispose to incident vertebral fractures in women, independent of BMD. The associations we observed were modest however, demonstrating the importance of conducting studies that are adequately powered to detect and quantify the effects of common genetic variants on complex diseases.

A road map for efficient and reliable human genome epidemiology

Networks of investigators have begun sharing best practices, tools and methods for analysis of associations between genetic variation and common diseases. A Network of Investigator Networks has been set up to drive the process, sponsored by the Human Genome Epidemiology Network. A workshop is planned to develop consensus guidelines for reporting results of genetic association studies. Published literature databases will be integrated, and unpublished data, including 'negative' studies, will be captured by online journals and through investigator networks. Systematic reviews will be expanded to include more meta-analyses of individual-level data and prospective meta-analyses. Field synopses will offer regularly updated overviews.

Selection bias in meta-analyses of gene-disease associations

Jin-Ling Tang discusses the issues arising from Ioannidis and colleagues' paper on biases in the Chinese genetic epidemiology literature, and possible solutions