Common outcome measures include relative risk, odds ratio, cor-
relation coefficient, risk difference, and (standardized) mean differ-
ence [5]. Since most of gene expression studies have small sample
sizes [6], meta-analysis is useful to increase statistical power by
combining multiple gene expression studies. A recent academic
literature review discussed pros and cons of genomic meta-analysis
for biomedical sciences, particularly on the identification of differ-
entially expressed genes, analysis of pathways/networks and predic-
tive models [6]. It is noticed that most of the genomic meta-
analysis methods are designed for those studies published with
the raw data. However, many of such genomic studies still do not
make their raw data readily accessible to the public. For this reason,
this chapter describes a proper statistical meta-analysis method
(instead of the simple vote-counting that was misused in many
genomic meta-analyses) that is widely used in evidence-based med-
icine [7] and applicable to the situations with or without raw data.
This chapter also covers the quality assessment (which are missing
from many genomic meta-analysis) of studies in accordance with
the evidence-based research guidelines including the PRISMA [8]
and MIAME [9]. Here, meta-analysis of microRNA (miRNA)
differential expression studies on type 2 diabetes (T2D) is taken
as an example.
Type 2 diabetes is a complex metabolic disorder characterized
by insulin resistance [10] that is often undetected until hyperglyce-
mia is observed [11]. The pathogenesis of T2D involves genetic,
environmental, and lifestyle factors. Over time, multiple organ
damages can occur, especially to the heart, blood vessels, eyes,
kidneys, and nerves [12]. According to the International Diabetes
Federation (IDF) [13], 415 million adults were estimated to live
with diabetes with 46.5% undiagnosed in 2014. By 2040, this
figure will rise to 642 million. High prevalence and no cure avail-
able make achieving and maintaining glycemic control as the pri-
mary goal of an initial pharmacological treatment to prevent
progressive deterioration, which can be markedly improved by
identifying novel early biomarkers and therapeutics for diabetes.
MiRNAs are likely candidates of the early biomarkers of T2D for
detecting and monitoring of the disease [14]. Meta-analysis may
help to screen the miRNA-related biomarkers of highest potential
for experimental and clinical validation.
MiRNAs are a class of small (approximately 22 nucleotides),
endogenous, noncoding, highly stable RNAs that regulate gene
and protein expression. Generally, precursors of miRNAs are
co-transcribed with their hosting gene in mammal [15] by RNA
polymerase II, which contain a characteristic stem-loop structure
[16]. The pri-miRNAs are subsequently cleaved by the nuclear
ribonuclease III (RNase III) enzyme Drosha to produce
pre-miRNAs of 70–100 nucleotides that are transported by
Exportin-5 into cytoplasm. Following the transportation, a further474 Hongmei Zhu and Siu-wai Leung