American Journal of Epidemiology Advance Access published December 11, 2014 American Journal of Epidemiology © The Author 2014. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail:
[email protected].
DOI: 10.1093/aje/kwu249
Human Genome Epidemiology (HuGE) Review A Systematic Appraisal of Field Synopses in Genetic Epidemiology: A HuGE Review
* Correspondence to Dr. Evangelos Evangelou, Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina 45110, Greece (e-mail:
[email protected]).
Initially submitted March 24, 2014; accepted for publication August 20, 2014.
Evidence from genetic association studies is accumulating rapidly. Field synopses have recently arisen as an unbiased way of systematically synthesizing this evidence. We performed a systematic review and appraisal of published field synopses in genetic epidemiology and assessed their main findings and methodological characteristics. We identified 61 eligible field synopses, published between January 1, 2007, and October 31, 2013, on 52 outcomes reporting 734 significant associations at the P < 0.05 level. The median odds ratio for these associations was 1.25 (interquartile range, 1.15–1.43). Egger’s test was the most common method (n = 30 synopses) of assessing publication bias. Only 12 synopses (20%) used the Venice criteria to evaluate the epidemiologic credibility of their findings (n = 449 variants). Eleven synopses (18%) were accompanied by an online database that has been regularly updated. These synopses received more citations (P = 0.01) and needed a larger research team (P = 0.02) than synopses without an online database. Overall, field synopses are becoming a valuable tool for the identification of common genetic variants, especially when researchers follow relevant methodological guidelines. Our work provides a summary of the current status of the field synopses published to date and may help interested readers efficiently identify the online resources containing the relevant genetic evidence. epidemiologic methods; field synopsis; genetic associations; genetic epidemiology; genome, human; meta-analysis
Abbreviations: GWAS, genome-wide association studies; HuGE, Human Genome Epidemiology; IQR, interquartile range.
Editor’s note: This article also appears on the website of the Human Genome Epidemiology Network (http://www.cdc. gov/genomics/hugenet/default.htm).
which has led to the adoption of stringent significance thresholds (5). A scientific enterprise with so many false-positive findings can lose credibility or impede progress in research and translation (6). Knowledge synthesis and integration of findings from candidate gene studies and GWAS came to be regarded as essential, especially after the recommendations of the Human Genome Epidemiology Network (HuGENet) were published (7, 8). Although similar approaches may have existed before, they became more apparent and standardized after 2007, when the first field synopses were published (9–12). Field synopses constitute a snapshot of the current state of knowledge on genetic associations in a particular field of research defined by a disease or trait (7) and provide a summary of the cumulative evidence on the genetic etiology of complex diseases and normal traits. Most importantly, one of their main aspects, according to the proposed framework, is the
The number of genetic association studies for complex traits has increased exponentially in the last several years (1, 2). As of November 2013, the Human Genome Epidemiology (HuGE) Navigator database (3) had cataloged more than 85,000 articles pertaining to both candidate gene studies and genome-wide association studies (GWAS). Despite their large number, the majority of the proposed candidate gene associations were not replicated in subsequent studies as a result of limited statistical power, as well as selective reporting and other biases (4). Moreover, the low prior probabilities for large numbers of comparisons among all candidate gene studies and in GWAS create the risk of false-positive findings, 1
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
Lazaros Belbasis, Orestis A. Panagiotou, Vasilios Dosis, and Evangelos Evangelou*
2 Belbasis et al.
evaluation of the epidemiologic credibility of their findings using various approaches. To date, several field synopses on various phenotypes have been published (10, 13, 14). However, to the best of our knowledge, no systematic overview of these studies has been performed. We conducted a systematic appraisal of the published field synopses, and here we present the results, as well as the reported findings and statistical and methodological approaches used in the eligible articles. METHODS
We searched PubMed (US National Library of Medicine, Bethesda, Maryland) for field synopses pertaining to genetic associations with a specific disease or trait that had been published between January 1, 2007, and October 31, 2013. We used the algorithm (“systematic review” OR meta-analysis OR “field synopsis” OR “umbrella review”) AND (gene OR genetic* OR polymorphism*). Following the HuGENet recommendations (7), we considered 2007 to be the year in which the first synopsis was published (10). Additionally, to ensure that we included all potentially eligible synopses, we complemented our PubMed search by searching HuGE Navigator (http://www.hugenavigator.net/), a regularly updated database focusing exclusively on genetic association studies and meta-analyses thereof (3), using the term “field synopsis.” Eligibility criteria
Eligible articles were considered on the basis of the definition adopted by Khoury et al. (7), which describes a field synopsis as a regularly updated snapshot of the current state of knowledge about genetic associations and common diseases. Ideally, a field synopsis should be freely available, should use online databases, should use objective and transparent criteria for grading the credibility of cumulative evidence, and should summarize and update the information in peer-reviewed articles on a regular basis (7). We considered as study units all field synopses that had provided a systematic review of genetic associations between a disease or normal trait and variants in any gene and had additionally synthesized the available data using meta-analytical techniques. We excluded reports that did not systematically review the literature; that a priori restricted their search to specific genes and/or polymorphisms; that did not pertain to DNA polymorphisms (e.g., RNA sequence); that focused on pharmacogenetics; that were limited to the presentation of technical characteristics of published databases of genetic associations; that focused exclusively on specific gene families or on multiple variants of a specific gene; or that appraised already published meta-analyses. Two investigators (L.B., V.D.) independently screened and identified eligible synopses. Disagreements were resolved by consensus after discussions with the other 2 investigators. Data extraction
The data were recorded from the main reports and from the supplementary material when needed. We extracted
Data analysis
We calculated the median numbers (and their interquartile ranges) of nominally significant variants, eligible studies, data sets, authors, and citations. We also calculated the median effect sizes (and their corresponding interquartile ranges) of all significant associations for binary outcomes. We accepted the threshold used by the field synopses’ authors to claim statistical significance. We converted the effect size so as to
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
Search strategy
information on all nonoverlapping nominally significant associations reported by the authors of the eligible papers in main analyses and subgroup analyses carried out in various ethnic groups. When more than 1 genetic model had been used for the analysis, we captured the significant associations from the most powerful model, using the following order: per-allele, additive, dominant, recessive, and other models. When a significant association was available in main and subgroup analyses, we extracted the data corresponding to the main analyses. If a significant association was present in multiple subgroup analyses but not in the main analysis, we included the association from the analysis with the greatest statistical power (largest sample size). When an association was reported as significant under both fixed-effects and random-effects meta-analysis models, we extracted the effect sizes corresponding to the random-effects model for associations with high heterogeneity and to the fixed-effects model for associations with low heterogeneity, as specified by the authors of the synopses. From each paper we extracted information on the first author, journal and year of publication, outcome, number of authors, main and subgroup analyses performed, number of eligible studies, total number of variants identified, number of polymorphisms claimed to be statistically significant, effect estimates and measures of uncertainty thereof (standard errors or confidence intervals), total sample size, and total number of data sets per significant association used for metaanalysis. We also recorded the criteria for performing metaanalysis, as well as the methods used (i.e., fixed-effects, random-effects, other). Additionally, we recorded the methods used for the assessment of between-study heterogeneity, small-study effects, publication bias, and/or other biases; the epidemiologic credibility of the identified associations (15); and adjustments for multiple comparisons (16). We also recorded whether evaluation of Hardy-Weinberg equilibrium had been performed; the threshold of statistical significance used by the authors to deem an identified association significant; the sensitivity analyses performed; and whether any GWAS had been included. We assumed that the significance threshold was set at 0.05 if the authors did not specify otherwise. We accessed the online “Catalog of Published GWAS” (http://www.genome.gov/gwastudies/) (17) to record whether there had been any published GWAS pertaining to the same outcomes as those in field synopses by the ending date of the synopses’ literature search and to determine whether the synopses’ authors had included all available GWAS data. Two investigators (L.B., V.D.) independently extracted the data. Consensus was reached in cases of disagreement among the authors.
Synopsis of Field Synopses 3
9,221 Articles Reviewed by Title Screening Reasons for Exclusion (n = 9,064) Meta-analyses of single variants or gene families (n = 3,357) Articles not evaluating genetic associations (n = 2,874) Single candidate gene studies or GWAS (n = 1,692) Nonresearch articles (n = 538) Articles not associated with humans (n = 288) Pharmacogenetic studies (n = 254) Articles without quantitative synthesis (n = 61)
Reasons for Exclusion (n = 87) Articles without quantitative synthesis (n = 24) Meta-analyses of single variants or gene families (n = 19) Nonresearch articles (n = 19) Articles not associated with humans (n = 12) Articles not evaluating genetic associations (n = 10) Articles appraising already published meta-analyses (n = 2) Pharmacogenetic studies (n = 1)
70 Articles Reviewed by Full-Text Screening
Reasons for Exclusion (n = 9) Meta-analyses of single variants or gene families (n = 5) Articles without quantitative synthesis (n = 3) Articles appraising already published meta-analyses (n = 1)
61 Eligible Field Synopses Published From January 1, 2007, to October 31, 2013
Figure 1. PubMed literature search used to locate and select field synopses of genetic association studies published between January 1, 2007, and October 31, 2013. GWAS, genome-wide association studies.
always reflect the risk allele for disease-related traits. We also calculated descriptive statistics for any sensitivity and subgroup analyses performed, and we summarized the methods and criteria used for meta-analysis and for assessment of publication bias/small-study effects. Finally, we estimated the number and percentage of associations that survived after the application of a more stringent statistical significance threshold. Whenever the published reports mentioned an accompanying online database, we accessed the website, verified the database’s existence, and assessed its frequency of data updates
and format. We checked whether the databases had been updated at least once in the subsequent 24 months after their creation and recorded the date of their last update. On December 28, 2013, we accessed ISI Web of Science (Thomson Reuters, New York, New York) to record the latest impact factor of the journals in which the field synopses had been published and the number of citations each field synopsis had received since its publication. We assessed whether the existence of a database affected the number of citations received and the target journal (based on the journal impact factor), using the Wilcoxon rank-sum test.
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
157 Articles Reviewed by Abstract Screening
4 Belbasis et al.
Table 1. Characteristics of Field Synopses of Genetic Association Studies Published Between January 1, 2007, and October 31, 2013 First Author, Year (Reference No.)
Outcome
No. of Eligible Articles
No. of Nominally Significant Associations
GWAS Used?
No. of Citations
Online Database (Website)
Abhary, 2009 (59)
Diabetic retinopathy
82
2
No
37
Alg, 2013 (28)
Intracranial aneurysms
66
19
Yes
0
No
Allen, 2008 (13)
Schizophrenia
1,179
24
No
464
Ariyaratnam, 2007 (9)
Ischemic stroke
60
5
No
60
Bertram, 2007 (10)
Alzheimer’s disease
789
24
No
653
Brouwer, 2010 (60)
Meningococcal disease
28
3
No
22
No
Brouwer, 2009 (21)
Pneumonococcal disease, meningococcal disease
76
2
No
72
No
Buurma, 2013 (61)
Preeclampsia
Calati, 2013 (39)
Borderline personality disorder
Castaldi, 2010 (62)
No SZGene (http://www. szgene.org/) No AlzGene (http://www. alzgene.org/)
7
No
1
No
28
0
No
0
No
COPD
108
4
No
39
COPDGene (http:// copdgene.org)
Chatzinasiou, 2011 (27)
Cutaneous melanoma
145
47
Yes
22
MelGene (http://www. melgene.org/)
Chen, 2012 (63)
Polypoidal choroidal vasculopathy
33
8
No
5
Debette, 2009 (64)
Cervical artery dissection
Dolan, 2010 (65)
Preterm birth
Dwyer, 2013 (66)
Vascular cognitive impairment
Gizer, 2009 (67)
Attention-deficit/hyperactivity disorder
No
7
1
No
25
No
61
5
No
20
PTBGene (http://ric. einstein.yu.edu/ ptbgene/)
104
3
No
0
No
8
No
231
No
NR
Gohil, 2009 (68)
Venous thromboembolism
173
10
No
56
No
Hamzi, 2011 (69)
Ischemic stroke
32
2
No
3
No
He, 2013 (40)
Henoch-Schonlein purpura, Henoch-Schonlein purpura nephritis
45
6
No
0
No
Jiang, 2012 (29)
Rheumatoid arthritis
216
43
Yes
2
DRAP (http://210.46.85. 180/DRAP/)
Jin, 2011 (70)
Hepatocellular carcinoma
26
6
No
8
No
Kitsios, 2007 (11)
Ischaemic heart failure
22
1
No
29
No
Lee, 2012 (71)
Fibromyalgia
18
1
No
14
No No
Li, 2010 (72)
Asthma
119
7
No
7
Lill, 2012 (14)
Parkinson’s disease
828
212
Yes
55
PDGene (http://www. pdgene.org/)
Loh, 2009 (73)
Gastric cancer
203
19
No
39
No
López-León, 2008 (74)
Major depressive disorder
183
5
No
122
No
Ma, 2014a (25)
Colorectal cancer
945
62
Yes
NA
No
Marjot, 2011 (75)
Cerebral venous thrombosis
26
2
No
10
No
Maxwell, 2011 (42)
Brain microbleeds
10
1
No
8
No
Mazaki, 2011 (76)
Pancreatic cancer
32
2
No
5
No Table continues
For this analysis, we excluded journals that are not indexed in ISI Web of Science. We examined whether the association between the number of citations and the existence of a database was altered when we adjusted for the number of months since publication, performing a generalized linear model analysis with the number of citations as the
dependent variable and the number of months since publication and the existence of a database as the independent variables. For recently published field synopses not yet indexed in ISI Web of Science, we performed a main analysis assuming that the number of citations these field synopses had received and the number of months since publication
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
542
Synopsis of Field Synopses 5
Table 1. Continued First Author, Year (Reference No.)
McColgan, 2010 (77)
Outcome
Intracranial aneurysms, subarachnoid hemorrhage
No. of Eligible Articles
No. of Nominally Significant Associations
30
3
GWAS Used?
No
No. of Citations
13
Online Database (Website)
No
Mooyaart, 2011 (78)
Diabetic nephropathy
132
24
No
34
No
Paternoster, 2010 (18)
Carotid intima-media thickness
95
1
No
15
No
Paternoster, 2009 (19)
White matter hyperintensities
45
1
No
25
No
Peck, 2008 (79)
Hemorrhagic stroke, subarachnoid hemorrhage, intracranial aneurysms
85
5
No
27
No
Metabolic syndrome Cerebral amyloid angiopathy
Rao, 2009 (81)
Ischemic stroke
Romero-Gomez, 2011 (82) Hepatitis C
88
5
No
21
No
136
1
No
1
No
NR
0
No
18
No
NR
1
No
17
No
Sava, 2014a (48)
Chronic lymphocytic leukemia
36
4
No
NA
No
Schild, 2013 (83)
Suicidal behavior
107
0
No
0
No
Seifuddin, 2012 (24)
Bipolar disorder
487
4
Yes
6
No
Smolonska, 2009 (84)
COPD
69
8
No
54
No
Srivastava, 2011 (85)
Gallbladder cancer
44
1
No
10
No
Staines-Urias, 2012 (23)
Preeclampsia
192
5
No
4
No
Taylor, 2013 (22)
Obsessive-compulsive disorder
113
3
No
1
No
Theodoratou, 2012 (26)
Colorectal cancer
635
42
Yes
11
Tüttelmann, 2007 (12)
Male infertility
7
2
No
94
No
Varvarigou, 2011 (86)
Obstructive sleep apnea
36
1
No
10
No
Vassos, 2014a (49)
Violence and aggression
185
0
No
NA
No
Wu, 2011 (87)
Cerebral palsy
11
1
No
8
No
Xin, 2009 (41)
Adult early-onset ischemic stroke
51
3
No
19
No
Yadav, 2013 (88)
Ischemic stroke
42
4
No
2
No
Zaffanello, 2011 (43)
Renal scar formation after urinary tract infections
18
2
No
6
No No
CRCgene (http://www.cphs. mvm.ed.ac.uk/projects/ CRCgene/)
Zdoukopoulos, 2008 (89)
Placental abruption
22
4
No
19
Zhang, 2013 (90)
Gestational diabetes mellitus
29
9
No
0
No
Zhang, 2011 (91)
Breast cancer
1,059
51
No
49
No
Zintzaras, 2010 (46)
Osteoarthritis
61
4
No
7
Zintzaras, 2009 (45)
Chronic lymphocytic leukemia
81
0
No
11
CUMAGAS-CLL (http:// biomath.med.uth.gr/)
Zintzaras, 2009 (47)
Peripheral arterial disease
41
4
No
21
CUMAGAS-PAD (http:// biomath.med.uth.gr/)
CUMAGAS-OSTEO (http://biomath.med.uth.gr/)
Abbreviations: COPD, chronic obstructive pulmonary disease; CUMAGAS, Cumulative Meta-Analysis of Genetic Association Studies; DRAP, Database of Rheumatoid Arthritis Related Polymorphisms; GWAS, genome-wide association studies; NA, not available; NR, not reported. a Article was published online (electronic publication) before November 2013; year refers to the year of print publication.
were both equal to zero and a sensitivity analysis excluding these synopses. Furthermore, we assessed whether the number of authors was higher in field synopses that were accompanied by an online database, using the Wilcoxon rank-sum test. We also examined the association between the existence of a database and the receipt of funding for
conducting the synopsis, using Fisher’s exact test. Additionally, to explore any potential factors influencing the existence of an online database, we performed a logistic regression analysis with the existence of a database as the dependent variable and the number of authors and the existence of funding as independent variables.
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
Povel, 2011 (80) Rannikmäe, 2013 (20)
6 Belbasis et al.
Statistical analyses were performed in STATA, version 12.0 (StataCorp LP, College Station, Texas). All P values presented are 2-tailed. RESULTS Eligible field synopses
Characteristics of published field synopses
Of the 61 eligible synopses, 58 pertained to binary outcomes and used the odds ratio as the effect-size metric; 1 synopsis (18) pertained to a continuous outcome and used standardized mean differences; and 2 synopses (19, 20) included both binary and continuous outcomes. Of the 58 synopses with binary outcomes, 55 used an α level of 0.05 and estimated 95% confidence intervals, 2 synopses (21, 22) used an α level of 0.01 and estimated 99% confidence intervals, and 1 synopsis (23) used an α level of 0.05 and estimated 99% confidence intervals. The median number of eligible studies across the included synopses was 68 (interquartile range (IQR), 32–143). In total, 734 associations were nominally significant based on the thresholds used by the authors of the synopses. Eighty-one associations, from 5 field synopses, were subjected to a more stringent threshold set by their authors using the Bonferroni correction, and 68 (84%) of them remained significant after this correction. The median number of nominally significant associations in each synopsis was 4 (IQR, 1–8), whereas the median number of data sets used for the meta-analyses of the 734 significant associations was 5 (IQR, 3–11). The median odds ratio for all nominally significant variants was 1.25 (IQR, 1.15– 1.43). GWAS were available at the time of completion of the literature search for 22 synopses, and 7 of them included these GWAS data in their analyses. These synopses pertained to bipolar disorder (24), colorectal cancer (25, 26), cutaneous melanoma (27), intracranial aneurysm (28), Parkinson’s disease (14), and rheumatoid arthritis (29). The decision on whether to conduct a meta-analysis (Table 2) was based on at least 1 of the following criteria: 1) the number of data sets available per variant (n = 30); 2) the number of subjects per polymorphism (n = 4); and 3) the minor allele frequency in healthy controls (n = 1). Three synopses used only fixed-effects models for synthesis of the available data, 24 synopses used only random-effects models, 33 synopses used both random-effects and fixed-effects models, and the authors of 1 synopsis did not report the meta-
Reporting on the credibility of the findings and adjustment for multiple comparisons
Twelve field synopses used the Venice criteria (15) to assess the epidemiologic credibility of the 449 variants included in the respective analyses, whereas 1 of them also used the Bayes factor (36) (212 variants). Only 155 of the 449 associations were found to have evidence of strong epidemiologic credibility (overall grade A) by application of the Venice criteria (15). Eight field synopses used at least 1 method to adjust for multiple comparisons. These methods included the Bonferroni correction (n = 5 synopses; 81 assessed variants), the falsepositive report probability (37) (n = 3 synopses; 64 assessed variants), or the Bayesian false-discovery probability (38) (n = 1 synopses; 42 assessed variants). Additionally, 3 field synopses evaluated the included studies using some quality score, including the Newcastle-Ottawa scale (n = 2) (39, 40) or other qualitative criteria defined by the research team (n = 1) (41). Moreover, 2 field synopses used the STrengthening the REporting of Genetic Association Studies (STREGA) statement, which aims to improve reporting in genetic association studies (42–44). More details are presented in Table 2.
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
Our literature search yielded 9,221 studies, of which 61 studies met the inclusion criteria, as shown in Figure 1. The eligible synopses were published between January 1, 2007, and October 31, 2013. These synopses evaluated genetic risk factors for 52 outcomes (Table 1). For 44 outcomes, there was only 1 field synopsis; for 1 outcome (ischemic stroke), there were 5 field synopses; for another outcome (intracranial aneurysms), there were 3 synopses; and for each of the remaining 6 outcomes (chronic lymphocytic leukemia, chronic obstructive pulmonary disease, colorectal cancer, meningococcal disease, preeclampsia, and subarachnoid hemorrhage), there were 2 synopses.
analysis method applied (Table 2). The most common criterion for performing a meta-analysis was the presence of at least 3 independent data sets per variant (n = 22), while 30 synopses did not apply any criteria for conducting a metaanalysis. Fifty-six synopses assessed heterogeneity, using Cochran’s Q statistic (n = 6), the I 2 statistic (n = 7), both Q and I 2 (n = 41), or Woolf’s test (n = 2). Thirty-eight field synopses evaluated Hardy-Weinberg equilibrium using P = 0.05 as the threshold (Table 2). To explore potential sources of heterogeneity, the authors of 34 field synopses performed a total of 47 subgroup analyses, whereas the authors of 29 synopses performed a total of 45 sensitivity analyses. Specifically, the subgroup analyses were based on ethnicity (n = 30 synopses), disease characteristics (n = 9), and population characteristics (n = 8). Similarly, the sensitivity analyses were based on the exclusion of studies violating Hardy-Weinberg equilibrium in controls (n = 14), the exclusion of the first published study (n = 9), sample size (n = 7), study characteristics (n = 5), the exclusion of the first positive study (n = 3), or an iterative exclusion of studies (n = 7) (Table 2). Five more field synopses excluded a priori all studies that were characterized by violation of Hardy-Weinberg equilibrium in the control group, as it was set by their inclusion criteria for the meta-analysis. Finally, the authors of 6 synopses performed a meta-regression to further explore potential sources of heterogeneity. A total of 47 synopses used 1 or more methods to evaluate publication bias. Evaluation with graphical tools included funnel plots (n = 25 synopses) and Galbraith plots (n = 1). The authors of 43 synopses performed at least 1 statistical test or applied at least 1 method for the assessment of publication bias and/or small-study effects, including Egger’s test (n = 30) (30), Begg’s test (n = 11) (31), Harbord’s test (n = 11) (32), the test for an excess of statistical significance (n = 5) (33), the fail-safe N method (n = 2) (34), or Duval and Tweedie’s trim-and-fill analysis (n = 1) (35) (Table 2).
Table 2. Methodological Details on Field Synopses of Genetic Association Studies Published Between January 1, 2007, and October 31, 2013
First Author, Year (Reference No.)
Meta-Analysis Model
Criteria for Conducting Meta-Analysis
Subgroup Analyses
Sensitivity Analyses
Evaluation of Heterogeneity
Assessment of HWE?
Evaluation of Credibility and Adjustment for Multiple Comparisons
Assessment of Publication Bias/ Small-Study Effects
Abhary, 2009 (59)
RE
≥5 data sets per variant
Ethnicity, disease characteristics
None
Qa
No
None
Funnel plots, Egger’s test
Alg, 2013 (28)
RE, FE
None
Ethnicity
Iterative exclusion of studies
Q, I 2
Yes
Bonferroni correction
Funnel plots, Egger’s test
Allen, 2008 (13)
RE, FE
Minor allele frequency ≥0.01 in healthy controls, ≥4 data sets per variant
Ethnicity
EFPS, HWEvb
Q, I 2
Yes
Venice criteria
Funnel plots, Harbord’s test, ESS test
Ariyaratnam, 2007 (9)
RE
≥3 different data sets per variant
Ethnicity
Sample size
Q, I 2
No
None
Funnel plots, Egger’s test
Bertram, 2007 (10)
RE
≥3 different data sets per variant
Ethnicity
EFPS, HWEv
Q, I 2
Yes
None
Begg’s test, Egger’s test
Brouwer, 2010 (60)
RE, FE
None
Disease characteristics
None
Q
No
None
None
RE, FE
None
None
None
Q
Yes
None
None
Buurma, 2013 (61)
RE
None
None
Sample size
I2
No
None
Funnel plots, Egger’s test, Begg’s test
Calati, 2013 (39)
RE, FE
≥3 different data sets per variant
None
None
Q, I 2
Yes
None
Funnel plots, Egger’s test
Castaldi, 2010 (62)
RE
≥3 different data sets per variant
Ethnicity, population EFPS, HWEv, study characteristics characteristics
Q, I 2
Yes
None
Egger’s test
Chatzinasiou, 2011 (27)
RE
≥4 different data sets per variant
Ethnicity
EFPS, HWEv
Q, I 2
Yes
Venice criteria
Harbord’s test, ESS test
Chen, 2012 (63)
RE, FE
None
None
None
Q
Yes
None
Egger’s test
Debette, 2009 (64)
FE
None
None
None
Q, I 2
No
None
None
Dolan, 2010 (65)
RE, FE
≥3 different data sets per variant
Ethnicity
EFPS, HWEv
Q, I 2
Yes
Venice criteria
None
Dwyer, 2013 (66)
RE, FE
≥3 different data sets per variant
Ethnicity
None
None
Yes
None
Egger’s test
Gizer, 2009 (67)
RE, FE
None
None
None
Q, I 2
No
None
Funnel plots
Gohil, 2009 (68)
RE
None
Ethnicity
None
Q, I 2
No
None
Funnel plots, Egger’s test
Hamzi, 2011 (69)
FE
None
None
None
None
No
None
None
He, 2013 (40)
RE, FE
None
None
None
None
No
None
Egger’s test
Jiang, 2012 (29)
RE, FE
None
Ethnicity
None
Q, I 2
No
None
Funnel plots, Egger’s test
Jin, 2011 (70)
RE, FE
≥3 different data sets per variant
None
None
Q, I
Yes
Venice criteria, FPRP
Egger’s test, Begg’s test Table continues
Synopsis of Field Synopses 7
Brouwer, 2009 (21)
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
First Author, Year (Reference No.)
Kitsios, 2007 (11)
Meta-Analysis Model
RE, FE
Criteria for Conducting Meta-Analysis
None
Subgroup Analyses
None
Sensitivity Analyses
HWEv
Evaluation of Heterogeneity
Assessment of HWE?
Evaluation of Credibility and Adjustment for Multiple Comparisons
Assessment of Publication Bias/ Small-Study Effects
Q, I 2
Yes
None
None
2
Yes
None
Funnel plots, Egger’s test
Lee, 2012 (71)
RE, FE
None
None
None
Q, I
Li, 2010 (72)
RE, FE
≥3 different data sets per variant
Population characteristics
None
Q, I 2
Yes
None
Funnel plots, Egger’s test, Begg’s test
Lill, 2012 (14)
RE
≥4 different data sets per variant
Ethnicity
EFPS, HWEv
I2
Yes
Bayes factor, Venice criteria
Harbord’s test
Loh, 2009 (73)
RE, FE
None
Ethnicity
None
Q, I 2
No
None
Funnel plots, Begg’s test
López-León, 2008 (74)
RE, FE
≥3 different data sets per variant
None
EFPS
I2
Yes
None
Funnel plots
Ma, 2014 (25)c
RE
≥3 different data sets per variant
Ethnicity
EFPS, exclusion of the first positive study, HWEv
Q, I 2
Yes
Venice criteria, FPRP, Bonferroni correction
Harbord’s test, ESS test
Marjot, 2011 (75)
RE
None
None
Iterative exclusion of studies
Q, I 2
No
None
Funnel plots, Egger’s test
Maxwell, 2011 (42)
FE
>100 subjects per variant
Ethnicity, population Sample size characteristics
Q, I 2
Yes
None
Fail-safe N method
Mazaki, 2011 (76)
RE, FE
None
Ethnicity, disease characteristics, population characteristics
None
Q, I 2
Yes
None
None
McColgan, 2010 (77)
RE, FE
None
None
Iterative exclusion of studies
Q
No
None
Funnel plots, Egger’s test
Mooyaart, 2011 (78)
RE
None
Ethnicity, disease characteristics
None
I2
No
None
Funnel plots, Egger’s test, Begg’s test
Paternoster, 2010 (18)
RE
≥3 different data sets per gene, >5,000 subjects per gene
Ethnicity, population Sample size characteristics
Q, I 2
Yes
None
Funnel plots
Paternoster, 2009 (19)
RE, FE
>2,000 subjects per None gene
None
Q, I 2
Yes
None
None
Peck, 2008 (79)
RE, FE
≥3 different data sets per variant
None
Q, I 2
No
None
Funnel plots, Egger’s test
Povel, 2011 (80)
RE
>4,000 subjects per Ethnicity variant, ≥3 different data sets per variant
None
I2
No
None
Funnel plots, Egger’s test, Begg’s test
Rannikmäe, 2013 (20)
RE, FE
None
Study characteristics
Q, I 2
No
None
Funnel plots, fail-safe N method
Disease characteristics
Ethnicity, disease characteristics
Table continues
8 Belbasis et al.
Table 2. Continued
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
Table 2. Continued
First Author, Year (Reference No.)
Meta-Analysis Model
Criteria for Conducting Meta-Analysis
Subgroup Analyses
Sensitivity Analyses
Evaluation of Heterogeneity
Assessment of HWE?
Evaluation of Credibility and Adjustment for Multiple Comparisons
Assessment of Publication Bias/ Small-Study Effects
Rao, 2009 (81)
RE, FE
≥3 different data sets per variant
None
Iterative exclusion of studies
Q
No
None
Funnel plots
Romero-Gomez, 2011 (82)
RE, FE
None
None
None
None
No
None
None
Sava, 2014 (48)c
RE, FE
None
None
None
Q, I 2
Yes
FPRP
Egger’s test
Schild, 2013 (83)
RE
None
Ethnicity
Iterative exclusion of studies
Q, I 2
No
None
Egger’s test, trim-and-fill analysis
Seifuddin, 2012 (24)
RE, FE
≥3 different data sets per variant
None
None
Woolf’s test
Yes
Bonferroni correction
None
Smolonska, 2009 (84)
RE, FE
≥3 different data sets per variant
Ethnicity
None
Woolf’s test
Yes
None
None
Srivastava, 2011 (85)
RE, FE
≥3 different data sets per variant
None
Iterative exclusion of studies
Q, I 2
Yes
Venice criteria
Egger’s test, Begg’s test
Staines-Urias, 2012 (23)
RE
≥3 different data sets per variant
Ethnicity
Exclusion of the first positive study, HWEv, sample size, study characteristics
Q, I 2
Yes
Venice criteria
Funnel plots, Galbraith plots, Harbord’s test, ESS test
Taylor, 2013 (22)
RE
≥5 different data sets per variant
None
None
I2
Yes
Bonferroni correction
Egger’s test
Theodoratou, 2012 (26)
RE, FE
≥4 different data sets per variant
Ethnicity
None
Q, I 2
No
Venice criteria, BFDP
Harbord’s test
NR
None
None
None
None
Yes
None
None
RE
≥3 different data sets per variant
Population characteristics
None
Q, I 2
Yes
None
Harbord’s test, Egger’s test
Vassos, 2014 (49)c
RE
≥3 different data sets per variant
Ethnicity, disease characteristics, population characteristics
Study characteristics
Q, I 2
No
Venice criteria
Begg’s test, Egger’s test
Wu, 2011 (87)
RE, FE
None
None
None
Q, I 2
Yes
Bonferroni correction
Funnel plots, Egger’s test
Xin, 2009 (41)
RE, FE
≥3 different data sets per variant
None
HWEv
Q, I 2
Yes
None
Begg’s test, Egger’s test
Yadav, 2013 (88)
RE
None
None
None
I2
No
None
Funnel plots, Egger’s test
Zaffanello, 2011 (43)
RE, FE
None
Ethnicity, disease characteristics, population characteristics
Study characteristics, sample size
Q, I 2
Yes
Venice criteria
None
Zdoukopoulos, 2008 (89)
RE, FE
None
None
None
Q, I 2
Yes
None
None
Zhang, 2013 (90)
RE
None
None
Iterative exclusion of studies
Q, I 2
Yes
None
Egger’s test Table continues
Synopsis of Field Synopses 9
Tüttelmann, 2007 (12) Varvarigou, 2011 (86)
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
Abbreviations: BFDP, Bayesian false-discovery probability; EFPS, exclusion of first published study; ESS, excess of statistical significance; FE, fixed effects; FPRP, false-positive report probability; HWE, Hardy-Weinberg equilibrium; HWEv, violation of Hardy-Weinberg equilibrium; NR, not reported; RE, random effects. a Cochran’s Q test. b Exclusion of studies violating Hardy-Weinberg equilibrium. c Article was published online (electronic publication) before November 2013; year refers to the year of print publication.
Harbord’s test Yes Ethnicity Zintzaras, 2009 (47)
RE
None
HWEv
Q, I 2
None
Harbord’s test Yes Ethnicity Zintzaras, 2009 (45)
RE
≥4 different data sets per variant
HWEv
Q, I 2
None
Harbord’s test Yes Ethnicity, disease characteristics Zintzaras, 2010 (46)
RE
≥4 different data sets per variant
HWEv
Q, I 2
None
Funnel plots, Begg’s test, Harbord’s test, ESS test Venice criteria Yes Q, I 2 EFPS, HWEv, exclusion of first positive study, sample size Ethnicity ≥3 different data sets per variant RE Zhang, 2011 (91)
Subgroup Analyses
Sensitivity Analyses
Evaluation of Heterogeneity Criteria for Conducting Meta-Analysis Meta-Analysis Model First Author, Year (Reference No.)
Table 2. Continued
Comparison of field synopses with overlapping outcomes
As we noted above, there were 18 synopses on 8 overlapping outcomes. With the exception of 2 synopses that evaluated genetic associations with ischemic stroke in South Asians and non-Europeans, all of the synopses considered participants of all ethnicities. The synopses on ischemic stroke, subarachnoid hemorrhage, meningococcal disease, and chronic obstructive pulmonary disease did not use any method to evaluate the credibility of the association or to adjust for multiple comparisons. Such tools were used in both studies on colorectal cancer, in 1 synopsis on chronic lymphocytic leukemia, and in 1 synopsis on preeclampsia. Assessment of publication bias or small-study effects was performed in most of the synopses (n = 14). In 5 of the 8 overlapping outcomes, not accounting for the 2 synopses limited to specific ethnicities, the last published synopsis did not have more eligible articles, suggesting that different search strategies and inclusion criteria may have been used. A detailed description of the synopses on overlapping outcomes is given in Appendix Table 1. Databases
Only 11 synopses (18%) were accompanied by online databases cataloging the findings and the studies included in the corresponding published field synopsis. In 4 databases (29, 45–47), the authors urged interested scientists to include their data in the database by granting access through a loggingin system. The median number of authors in the field synopses supported by an online database was 12, while in field synopses without an online database the median number of authors was 4 (P = 0.02, Wilcoxon rank-sum test). Funding was associated with the existence of an online database (P = 0.046, Fisher’s exact test). However, when the number of authors and the existence of funding were considered in a logistic regression model, the existence of a database was affected only by the number of authors (P = 0.007) and not by the existence of funding (P = 0.238). All tracked databases are available online and can be accessed through their initial Web address as provided by the authors. Nine of the 11 databases have been online at least 24 months, and all of them were updated at least once in the 24 months after their creation, whereas 2 databases (14, 29) have been online for less than 24 months and therefore have not been evaluated with regard to an update. Citation analysis
Three field synopses (25, 48, 49) were published recently and have not yet been indexed in ISI Web of Science. The remaining 58 synopses have received 2,512 citations in total, with a median of 13 citations (IQR, 4–29). Field synopses supported by an online database received a median of 21 citations, while those without a database received a median of 10 citations (P = 0.073, Wilcoxon ranksum test). However, this difference was significant when we adjusted for number of months since publication (P = 0.007) using generalized linear regression. There were no significant differences in the median journal impact factor between
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
Assessment of HWE?
Evaluation of Credibility and Adjustment for Multiple Comparisons
Assessment of Publication Bias/ Small-Study Effects
10 Belbasis et al.
Synopsis of Field Synopses 11
synopses with (7.692) and without (5.419) a database (P = 0.157). The results were similar in a sensitivity analysis where we excluded nonindexed field synopses. DISCUSSION
ACKNOWLEDGMENTS
Author affiliations: Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina, Greece (Lazaros Belbasis, Vasilios Dosis, Evangelos Evangelou); Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, Maryland (Orestis A. Panagiotou); and Department of Epidemiology and Biostatistics, Imperial College London, London, United Kingdom (Evangelos Evangelou).
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
To our knowledge, this is the first systematic appraisal of field synopses in genetic epidemiology summarizing the current status of the synopses published to date, along with their related online resources whenever available. We identified 61 published field synopses covering a wide spectrum of outcomes and more than 700 variants that were deemed nominally significant. However, only about one-third of those variants showed strong evidence of credibility for the respective associations whenever such criteria were applied. The evidence of genetic associations was mostly assessed through the Venice criteria, the Bayes factor, or other statisticsbased approaches to adjustment for multiple comparisons, such as the Bonferroni correction, the Bayesian false-discovery probability, and the false-positive report probability. In most cases, the credibility of the findings was assessed and interpreted efficiently, providing information on the variants that had weak, moderate, or strong epidemiologic credibility. However, more than half of the published field synopses failed to either provide any assessment for their associations or apply a stringent P-value threshold, and therefore these findings should be interpreted with caution to avoid falsepositive inferences. At the same time, given that the stringent genome-wide significance thresholds applied in GWAS may increase the number of false-negative findings, it is important for researchers to consider the balance between false-positive and false-negative findings (50). Regarding dissemination of the results, we found that only about one-sixth of the synopses were supported by an online database as previously suggested (7), the presence of which improved the impact and visibility of the article. Synopses accompanied by an online database seem to attract significantly more citations than synopses without a database. All available databases have been updated and have incorporated new features for data-searching and analysis over time, improving the experience for the involved investigators and clinicians and promoting the relevant scientific fields. As expected, all field synopses used meta-analytical techniques to synthesize available evidence. Meta-analysis is now a sine qua non for the discovery of common variants for complex diseases (51); it is especially valuable when applied to studies retrieved through systematic literature reviews, such as those typically performed in field synopses. In this context, meta-analytical techniques and related “diagnostic” procedures can help identify biases (e.g., small-study effects, excess significance bias, and other types of bias) and assess between-study heterogeneity and its potential sources, and in general they provide insights about the validity of the proposed associations. Overall, the current findings support the crucial role of meta-analysis in the discovery of novel genetic associations either through synthesis of published studies or through meta-analysis of individual-level data as routinely performed in GWAS consortia (52). Our study has some caveats that should be acknowledged. First, we did not include synopses focusing on gene families,
gene pathways (53), or multiple variants of a specific gene (54), even though such approaches may have been conducted rigorously. In addition, we did not include databases that were not accompanied by published articles describing the results in detail or a published article focusing on technical characteristics, such as MSGene (http://www.msgene.org/), ALSGene (55), or CADgene (56). These initiatives are expected to follow stringent criteria for the synthesis and reporting of the available data, and they add to the existing evidence of genetic associations. Such efforts should be encouraged and should follow the proper guidelines as previously reported (7, 15, 44). Furthermore, some investigators have systematically appraised the published meta-analyses of genetic association studies related to a specific disease. For example, in an overview of meta-analyses, Dong et al. (57) examined the validity of genetic associations for cancer, estimating the false-positive report probability. This methodology did not provide a complete snapshot of all published genetic associations related to cancer susceptibility, but it contributed to the identification of false-positive genetic associations. Moreover, we did not capture articles published before 2007, given that the actual definition of a field synopsis and the ideal description of a field synopsis had not been set by that time. Finally, we stress that our work summarizes and provides a snapshot of the field without assessing the methodological approaches used by the synopses’ authors. Of course, all of the approaches may have certain disadvantages, and it is possible that in some cases they may have been used inappropriately. Field synopses have been proven to be a valuable tool for identifying common variants in the past, and we expect that new findings will be continuously added in the future. This method could also be proven useful for rare variants, where single studies will mostly lack the statistical power to show a significant association. Therefore, it is obvious that the rapid evolution of the field necessitates such systematic approaches for the quantitative synthesis and critical appraisal of significant associations. However, researchers should adhere to rules and guidelines, such as those provided by the HuGENet road map (7), in order to achieve adequate quality and reporting of the identified signals. Similar approaches, such as umbrella reviews, are already being used in other fields of epidemiology to systematically appraise the validity and credibility of different factors, including biomarkers or other environmental risk factors (58). The existence of an online database and the application of statistical approaches to evaluate the cumulative evidence and to adjust for multiple comparisons may improve the reporting and impact of field synopses.
12 Belbasis et al.
The findings and conclusions in this report are those of the authors and do not necessarily represent the official views of the National Cancer Institute or the US Department of Health and Human Services. Conflict of interest: none declared.
18.
19. REFERENCES 20.
21.
22. 23. 24. 25.
26.
27.
28. 29.
30. 31. 32. 33. 34. 35. 36.
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
1. Ioannidis JP. Expectations, validity, and reality in omics. J Clin Epidemiol. 2010;63(9):945–949. 2. Lin BK, Clyne M, Walsh M, et al. Tracking the epidemiology of human genes in the literature: the HuGE Published Literature database. Am J Epidemiol. 2006;164(1):1–4. 3. Yu W, Gwinn M, Clyne M, et al. A navigator for human genome epidemiology. Nat Genet. 2008;40(2):124–125. 4. Ioannidis JP, Ntzani EE, Trikalinos TA, et al. Replication validity of genetic association studies. Nat Genet. 2001;29(3): 306–309. 5. Panagiotou OA, Ioannidis JP; Genome-Wide Significance Project. What should the genome-wide significance threshold be? Empirical replication of borderline genetic associations. Int J Epidemiol. 2012;41(1):273–286. 6. Khoury MJ, Gwinn M, Ioannidis JP. The emergence of translational epidemiology: from scientific discovery to population health impact. Am J Epidemiol. 2010;172(5): 517–524. 7. Khoury MJ, Bertram L, Boffetta P, et al. Genome-wide association studies, field synopses, and the development of the knowledge base on genetic variation and human diseases. Am J Epidemiol. 2009;170(3):269–279. 8. Ioannidis JP, Gwinn M, Little J, et al. A road map for efficient and reliable human genome epidemiology. Nat Genet. 2006; 38(1):3–5. 9. Ariyaratnam R, Casas JP, Whittaker J, et al. Genetics of ischaemic stroke among persons of non-European descent: a meta-analysis of eight genes involving approximately 32,500 individuals. PLoS Med. 2007;4(4):e131. 10. Bertram L, McQueen MB, Mullin K, et al. Systematic meta-analyses of Alzheimer disease genetic association studies: the AlzGene database. Nat Genet. 2007;39(1):17–23. 11. Kitsios G, Zintzaras E. Genetic variation associated with ischemic heart failure: a HuGE review and meta-analysis. Am J Epidemiol. 2007;166(6):619–633. 12. Tüttelmann F, Rajpert-De Meyts E, Nieschlag E, et al. Gene polymorphisms and male infertility—a meta-analysis and literature review. Reprod Biomed Online. 2007;15(6):643–658. 13. Allen NC, Bagade S, McQueen MB, et al. Systematic meta-analyses and field synopsis of genetic association studies in schizophrenia: the SzGene database. Nat Genet. 2008;40(7): 827–834. 14. Lill CM, Roehr JT, McQueen MB, et al. Comprehensive research synopsis and systematic meta-analyses in Parkinson’s disease genetics: the PDGene database. PLoS Genet. 2012;8(3): e1002548. 15. Ioannidis JP, Boffetta P, Little J, et al. Assessment of cumulative evidence on genetic associations: interim guidelines. Int J Epidemiol. 2008;37(1):120–132. 16. Attia J, Ioannidis JP, Thakkinstian A, et al. How to use an article about genetic association: B: are the results of the study valid? JAMA. 2009;301(2):191–197. 17. Hindorff LA, Sethupathy P, Junkins HA, et al. Potential etiologic and functional implications of genome-wide
association loci for human diseases and traits. Proc Natl Acad Sci U S A. 2009;106(23):9362–9367. Paternoster L, Martinez-Gonzalez NA, Charleton R, et al. Genetic effects on carotid intima-media thickness: systematic assessment and meta-analyses of candidate gene polymorphisms studied in more than 5000 subjects. Circ Cardiovasc Genet. 2010;3(1):15–21. Paternoster L, Chen W, Sudlow CL. Genetic determinants of white matter hyperintensities on brain scans: a systematic assessment of 19 candidate gene polymorphisms in 46 studies in 19,000 subjects. Stroke. 2009;40(6):2020–2026. Rannikmäe K, Samarasekera N, Martînez-Gonzâlez NA, et al. Genetics of cerebral amyloid angiopathy: systematic review and meta-analysis. J Neurol Neurosurg Psychiatry. 2013;84(8): 901–908. Brouwer MC, de Gans J, Heckenberg SG, et al. Host genetic susceptibility to pneumococcal and meningococcal disease: a systematic review and meta-analysis. Lancet Infect Dis. 2009; 9(1):31–44. Taylor S. Molecular genetics of obsessive-compulsive disorder: a comprehensive meta-analysis of genetic association studies. Mol Psychiatry. 2013;18(7):799–805. Staines-Urias E, Paez MC, Doyle P, et al. Genetic association studies in pre-eclampsia: systematic meta-analyses and field synopsis. Int J Epidemiol. 2012;41(6):1764–1775. Seifuddin F, Mahon PB, Judy J, et al. Meta-analysis of genetic association studies on bipolar disorder. Am J Med Genet B Neuropsychiatr Genet. 2012;159B(5):508–518. Ma X, Zhang B, Zheng W. Genetic variants associated with colorectal cancer risk: comprehensive research synopsis, meta-analysis, and epidemiological evidence. Gut. 2014;63(2): 326–336. Theodoratou E, Montazeri Z, Hawken S, et al. Systematic meta-analyses and field synopsis of genetic association studies in colorectal cancer. J Natl Cancer Inst. 2012;104(19): 1433–1457. Chatzinasiou F, Lill CM, Kypreou K, et al. Comprehensive field synopsis and systematic meta-analyses of genetic association studies in cutaneous melanoma. J Natl Cancer Inst. 2011; 103(16):1227–1235. Alg VS, Sofat R, Houlden H, et al. Genetic risk factors for intracranial aneurysms: a meta-analysis in more than 116,000 individuals. Neurology. 2013;80(23):2154–2165. Jiang Y, Zhang R, Zheng J, et al. Meta-analysis of 125 rheumatoid arthritis-related single nucleotide polymorphisms studied in the past two decades. PLoS One. 2012;7(12):e51571. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997; 315(7109):629–634. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics. 1994;50(4): 1088–1101. Harbord RM, Egger M, Sterne JA. A modified test for small-study effects in meta-analyses of controlled trials with binary endpoints. Stat Med. 2006;25(20):3443–3457. Ioannidis JP, Trikalinos TA. An exploratory test for an excess of significant findings. Clin Trials. 2007;4(3):245–253. Rosenthal R. The file drawer problem and tolerance for null results. Psychol Bull. 1979;86(3):638–641. Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455–463. Goodman SN. Toward evidence-based medical statistics. 2: the Bayes factor. Ann Intern Med. 1999;130(12):1005–1013.
Synopsis of Field Synopses 13
56. Liu H, Liu W, Liao Y, et al. CADgene: a comprehensive database for coronary artery disease genes. Nucleic Acids Res. 2011;39(database issue):D991–D996. 57. Dong LM, Potter JD, White E, et al. Genetic susceptibility to cancer: the role of polymorphisms in candidate genes. JAMA. 2008;299(20):2423–2436. 58. Ioannidis JP. Integration of evidence from multiple meta-analyses: a primer on umbrella reviews, treatment networks and multiple treatments meta-analyses. CMAJ. 2009; 181(8):488–493. 59. Abhary S, Hewitt AW, Burdon KP, et al. A systematic meta-analysis of genetic association studies for diabetic retinopathy. Diabetes. 2009;58(9):2137–2147. 60. Brouwer MC, Read RC, van de Beek D. Host genetics and outcome in meningococcal disease: a systematic review and meta-analysis. Lancet Infect Dis. 2010;10(4):262–274. 61. Buurma AJ, Turner RJ, Driessen JH, et al. Genetic variants in pre-eclampsia: a meta-analysis. Hum Reprod Update. 2013; 19(3):289–303. 62. Castaldi PJ, Cho MH, Cohn M, et al. The COPD genetic association compendium: a comprehensive online database of COPD genetic associations. Hum Mol Genet. 2010;19(3): 526–534. 63. Chen H, Liu K, Chen LJ, et al. Genetic associations in polypoidal choroidal vasculopathy: a systematic review and meta-analysis. Mol Vis. 2012;18:816–829. 64. Debette S, Markus HS. The genetics of cervical artery dissection: a systematic review. Stroke. 2009;40(6):e459–e466. 65. Dolan SM, Hollegaard MV, Merialdi M, et al. Synopsis of preterm birth genetic association studies: the Preterm Birth Genetics Knowledge Base (PTBGene). Public Health Genomics. 2010;13(7-8):514–523. 66. Dwyer R, Skrobot OA, Dwyer J, et al. Using Alzgene-like approaches to investigate susceptibility genes for vascular cognitive impairment. J Alzheimers Dis. 2013;34(1):145–154. 67. Gizer IR, Ficks C, Waldman ID. Candidate gene studies of ADHD: a meta-analytic review. Hum Genet. 2009;126(1): 51–90. 68. Gohil R, Peck G, Sharma P. The genetics of venous thromboembolism. A meta-analysis involving approximately 120,000 cases and 180,000 controls. Thromb Haemost. 2009; 102(2):360–370. 69. Hamzi K, Tazzite A, Nadifi S. Large-scale meta-analysis of genetic studies in ischemic stroke: five genes involving 152,797 individuals. Indian J Hum Genet. 2011;17(3): 212–217. 70. Jin F, Xiong WJ, Jing JC, et al. Evaluation of the association studies of single nucleotide polymorphisms and hepatocellular carcinoma: a systematic review. J Cancer Res Clin Oncol. 2011; 137(7):1095–1104. 71. Lee YH, Choi SJ, Ji JD, et al. Candidate gene studies of fibromyalgia: a systematic review and meta-analysis. Rheumatol Int. 2012;32(2):417–426. 72. Li X, Zhang Y, Zhang J, et al. Asthma susceptible genes in Chinese population: a meta-analysis. Respir Res. 2010;11: 129–150. 73. Loh M, Koh KX, Yeo BH, et al. Meta-analysis of genetic polymorphisms and gastric cancer risk: variability in associations according to race. Eur J Cancer. 2009;45(14): 2562–2568. 74. López-León S, Janssens AC, González-Zuloeta Ladd AM, et al. Meta-analyses of genetic studies on major depressive disorder. Mol Psychiatry. 2008;13(8):772–785. 75. Marjot T, Yadav S, Hasan N, et al. Genes associated with adult cerebral venous thrombosis. Stroke. 2011;42(4):913–918.
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
37. Wacholder S, Chanock S, Garcia-Closas M, et al. Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J Natl Cancer Inst. 2004; 96(6):434–442. 38. Wakefield J. A Bayesian measure of the probability of false discovery in genetic epidemiology studies. Am J Hum Genet. 2007;81(2):208–227. 39. Calati R, Gressier F, Balestri M, et al. Genetic modulation of borderline personality disorder: systematic review and meta-analysis. J Psychiatr Res. 2013;47(10):1275–1287. 40. He X, Yu C, Zhao P, et al. The genetics of Henoch-Schönlein purpura: a systematic review and meta-analysis. Rheumatol Int. 2013;33(6):1387–1395. 41. Xin XY, Song YY, Ma JF, et al. Gene polymorphisms and risk of adult early-onset ischemic stroke: a meta-analysis. Thromb Res. 2009;124(5):619–624. 42. Maxwell SS, Jackson CA, Paternoster L, et al. Genetic associations with brain microbleeds: systematic review and meta-analyses. Neurology. 2011;77(2):158–167. 43. Zaffanello M, Tardivo S, Cataldi L, et al. Genetic susceptibility to renal scar formation after urinary tract infection: a systematic review and meta-analysis of candidate gene polymorphisms. Pediatr Nephrol. 2011;26(7):1017–1029. 44. Little J, Higgins JP, Ioannidis JP, et al. STrengthening the REporting of Genetic Association Studies (STREGA): an extension of the STROBE Statement. PLoS Med. 2009;6(2): e22. 45. Zintzaras E, Kitsios GD. Synopsis and synthesis of candidate-gene association studies in chronic lymphocytic leukemia: the CUMAGAS-CLL information system. Am J Epidemiol. 2009;170(6):671–678. 46. Zintzaras E, Kitsios GD, Ziogas DC, et al. Field synopsis and synthesis of genetic association studies in osteoarthritis: the CUMAGAS-OSTEO information system. Am J Epidemiol. 2010;171(8):851–858. 47. Zintzaras E, Zdoukopoulos N. A field synopsis and meta-analysis of genetic association studies in peripheral arterial disease: the CUMAGAS-PAD database. Am J Epidemiol. 2009;170(1):1–11. 48. Sava GP, Speedy HE, Houlston RS. Candidate gene association studies and risk of chronic lymphocytic leukemia: a systematic review and meta-analysis. Leuk Lymphoma. 2014;55(1): 160–167. 49. Vassos E, Collier DA, Fazel S. Systematic meta-analyses and field synopsis of genetic association studies of violence and aggression. Mol Psychiatry. 2014;19(4):471–477. 50. Ioannidis JP, Tarone R, McLaughlin JK. The false-positive to false-negative ratio in epidemiologic studies. Epidemiology. 2011;22(4):450–456. 51. Evangelou E, Ioannidis JP. Meta-analysis methods for genome-wide association studies and beyond. Nat Rev Genet. 2013;14(6):379–389. 52. Panagiotou OA, Willer CJ, Hirschhorn JN, et al. The power of meta-analysis in genome-wide association studies. Annu Rev Genomics Hum Genet. 2013;14:441–465. 53. Vineis P, Manuguerra M, Kavvoura FK, et al. A field synopsis on low-penetrance variants in DNA repair genes and cancer susceptibility. J Natl Cancer Inst. 2009;101(1):24–36. 54. Gao J, Pan Z, Jiao Z, et al. TPH2 gene polymorphisms and major depression—a meta-analysis. PLoS One. 2012;7(5): e36721. 55. Lill CM, Abel O, Bertram L, et al. Keeping up with genetic discoveries in amyotrophic lateral sclerosis: the ALSoD and ALSGene databases. Amyotroph Lateral Scler. 2011;12(4): 238–249.
14 Belbasis et al.
84. Smolonska J, Wijmenga C, Postma DS, et al. Meta-analyses on suspected chronic obstructive pulmonary disease genes: a summary of 20 years’ research. Am J Respir Crit Care Med. 2009;180(7):618–631. 85. Srivastava K, Srivastava A, Sharma KL, et al. Candidate gene studies in gallbladder cancer: a systematic review and meta-analysis. Mutat Res. 2011;728(1-2):67–79. 86. Varvarigou V, Dahabreh IJ, Malhotra A, et al. A review of genetic association studies of obstructive sleep apnea: field synopsis and meta-analysis. Sleep. 2011;34(11): 1461–1468. 87. Wu D, Zou YF, Xu XY, et al. The association of genetic polymorphisms with cerebral palsy: a meta-analysis. Dev Med Child Neurol. 2011;53(3):217–225. 88. Yadav S, Hasan N, Marjot T, et al. Detailed analysis of gene polymorphisms associated with ischemic stroke in South Asians. PLoS One. 2013;8(3):e57305. 89. Zdoukopoulos N, Zintzaras E. Genetic risk factors for placental abruption: a HuGE review and meta-analysis. Epidemiology. 2008;19(2):309–323. 90. Zhang C, Bao W, Rong Y, et al. Genetic variants and the risk of gestational diabetes mellitus: a systematic review. Hum Reprod Update. 2013;19(4):376–390. 91. Zhang B, Beeghly-Fadiel A, Long J, et al. Genetic variants associated with breast-cancer risk: comprehensive research synopsis, meta-analysis, and epidemiological evidence. Lancet Oncol. 2011;12(5):477–488.
(Appendix follows)
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
76. Mazaki T, Masuda H, Takayama T. Polymorphisms and pancreatic cancer risk: a meta-analysis. Eur J Cancer Prev. 2011;20(3):169–183. 77. McColgan P, Thant KZ, Sharma P. The genetics of sporadic ruptured and unruptured intracranial aneurysms: a genetic meta-analysis of 8 genes and 13 polymorphisms in approximately 20,000 individuals. J Neurosurg. 2010;112(4): 714–721. 78. Mooyaart AL, Valk EJ, van Es LA, et al. Genetic associations in diabetic nephropathy: a meta-analysis. Diabetologia. 2011; 54(3):544–553. 79. Peck G, Smeeth L, Whittaker J, et al. The genetics of primary haemorrhagic stroke, subarachnoid haemorrhage and ruptured intracranial aneurysms in adults. PLoS One. 2008;3(11):e3691. 80. Povel CM, Boer JM, Reiling E, et al. Genetic variants and the metabolic syndrome: a systematic review. Obes Rev. 2011; 12(11):952–967. 81. Rao R, Tah V, Casas JP, et al. Ischaemic stroke subtypes and their genetic basis: a comprehensive meta-analysis of small and large vessel stroke. Eur Neurol. 2009;61(2):76–86. 82. Romero-Gomez M, Eslam M, Ruiz A, et al. Genes and hepatitis C: susceptibility, fibrosis progression and response to treatment. Liver Int. 2011;31(4):443–460. 83. Schild AH, Pietschnig J, Tran US, et al. Genetic association studies between SNPs and suicidal behavior: a meta-analytical field synopsis. Prog Neuropsychopharmacol Biol Psychiatry. 2013;46:36–42.
Appendix Table 1. Characteristics of Field Synopses of Genetic Association Studies With Overlapping Outcomes, January 1, 2007–October 31, 2013 First Author, Year (Reference No.)
Date of Last Literature Search
Database(s)a Searched
Racial/Ethnic Population Covered
No. of Eligible Articles
No. of Nominally Significant Associations
Criteria for Conducting Meta-Analysis
Evaluation of Credibility and Adjustment for Multiple Comparisons
Assessment of Publication Bias/ Small-Study Effects
Ischemic Stroke Yadav, 2013 (88)
August 2012
PubMed, EMBASE, Google Scholar
South Asians
42
4
None
None
Funnel plots, Egger’s test
Hamzi, 2011 (69)
January 2010
PubMed, EMBASE, Scopus
All ethnicities
32
2
None
None
None
Rao, 2009 (81)
January 2007
PubMed, EMBASE, Google Scholar
All ethnicities
NR
0
≥3 different data sets per variant
None
Funnel plots
Xin, 2009b (41)
December 2008 PubMed, EMBASE, Google Scholar
All ethnicities
51
3
≥3 different data sets per variant
None
Begg’s test, Egger’s test
Ariyaratnam, 2007 (9)
January 2005
Non-Europeans
60
5
≥3 different data sets per variant
None
Funnel plots, Egger’s test
Alg, 2013 (28)
December 2012 PubMed, EMBASE, Google Scholar
All ethnicities
66
19
None
Bonferroni correction
Funnel plots, Egger’s test
McColgan, 2010 (77)
June 2008
PubMed, EMBASE, Google Scholar, Yahoo!
All ethnicities
30
3
None
None
Funnel plots, Egger’s test
Peck, 2008 (79)
March 2007
PubMed, EMBASE, Google Scholar
All ethnicities
85
5
≥3 different data sets per variant
None
Funnel plots, Egger’s test
Sava, 2014c (48)
October 2012
PubMed
All ethnicities
36
4
None
FPRP
Egger’s test
Zintzaras, 2009 (45)
February 2009
PubMed, HuGE Navigator
All ethnicities
81
0
≥4 different data sets per variant
None
Harbord’s test
Castaldi, 2010 (62)
July 2008
PubMed, HuGE Navigator
All ethnicities
108
4
≥3 different data sets per variant
None
Egger’s test
Smolonska, 2009 (84)
NR
PubMed
All ethnicities
69
8
≥3 different data sets per variant
None
None
PubMed, EMBASE, Google Scholar, Yahoo!
Intracranial Aneurysms
Chronic Lymphocytic Leukemia
Colorectal Cancer Ma, 2014 (25)
December 2012 PubMed, Google Scholar
Theodoratou, 2012 (26) March 2011
PubMed, HuGE Navigator
All ethnicities
945
62
≥3 different data sets per variant
Venice criteria, FPRP, Bonferroni correction
Harbord’s test, ESS test
All ethnicities
635
42
≥4 different data sets per variant
Venice criteria, BFDP
Harbord’s test Table continues
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014
c
Synopsis of Field Synopses 15
Chronic Obstructive Pulmonary Disease
16 Belbasis et al.
Appendix Table 1. Continued First Author, Year (Reference No.)
Date of Last Literature Search
Database(s)a Searched
Racial/Ethnic Population Covered
No. of Eligible Articles
No. of Nominally Significant Associations
Criteria for Conducting Meta-Analysis
Evaluation of Credibility and Adjustment for Multiple Comparisons
Assessment of Publication Bias/ Small-Study Effects
Meningococcal Disease Brouwer, 2010 (60)
October 2009
PubMed
All ethnicities
28
3
None
None
None
Brouwer, 2009 (21)
August 2008
PubMed
All ethnicities
76
2
None
None
None
Preeclampsia Buurma, 2013 (61)
February 2012
PubMed, EMBASE, ISI Web of Science
All ethnicities
542
7
None
None
Funnel plots, Egger’s test, Begg’s test
Staines-Urias, 2012 (23)
January 2011
PubMed, EMBASE, KoreaMed
All ethnicities
192
5
≥3 different data sets per variant
Venice criteria
Funnel plots, Galbraith plots, Harbord’s test, ESS test
McColgan, 2010 (77)
June 2008
PubMed, EMBASE, Google Scholar, Yahoo!
All ethnicities
30
3
None
None
Funnel plots, Egger’s test
Peck, 2008 (79)
March 2007
PubMed, EMBASE, Google Scholar
All ethnicities
85
5
≥3 different data sets per variant
None
Funnel plots, Egger’s test
Subarachnoid Hemorrhage
Abbreviations: BFDP, Bayesian false-discovery probability; EMBASE, Excerpta Medica Database; ESS, excess of statistical significance; FPRP, false-positive report probability; HuGE, Human Genome Epidemiology; ISI, Institute for Scientific Information; NR, not reported. a PubMed: US National Library of Medicine, Bethesda, Maryland; ISI Web of Science: Thomson Reuters, New York, New York; Google Scholar: Google Inc., Mountain View, California; EMBASE: Elsevier B.V., Amsterdam, the Netherlands; Scopus: Elsevier B.V.; Yahoo!: Yahoo! Inc., Sunnyvale, California; HuGE Navigator: Centers for Disease Control and Prevention, Atlanta, Georgia; KoreaMed: Korean Association of Medical Journal Editors, Seoul, South Korea. b The outcome was defined as adult early-onset ischemic stroke. c Article was published online (electronic publication) before November 2013; year refers to the year of print publication.
Downloaded from http://aje.oxfordjournals.org/ at University of California, San Francisco on December 12, 2014