
Data collection Data extraction was performed independently
by two reviewers in a predefined database. The following data
were extracted: Author, year, country, study design, sample,
age, ethnicity of the sample (% for each ethnic group (skin
color or origin of population)), proportion of males and fe-
males for each sample, calculation of statistical power, cate-
gorization of dental caries, minor allele frequency calculation
for each polymorphism, analytical approach, Hardy-
Weinberg equilibrium, effect estimate (crude an d adjusted
analysis values and their respective confidence intervals)
covariables, and the main results.
Quality of studies The quality of the included studies was
verified in accordance with the Appraisal Checklist for
Observational Studies scale (Joanna Briggs Institute) (T.J.B.,
2014). This tool presents 10 questions evaluating different
points in the study, which should be answered with “No,”
“not clear,” or “Yes.” Each Yes response corresponds to one
point, so the tool score ranges from 0 to 10. Studies totaling
between 0 and 3 points were considered low quality; 4 to 6
were of medium quality; and 7 to 10 were considered high
quality. To classify the studies, two reviewers will perform the
classification independently. Disagreements will be remedied
through discussion until consensus is reached.
Strategy for data synthesis A meta-analysis was planned to be
performed when the same polymorphisms were identified in
at least three different studies, when the effects were shown or
where it was p ossible to calculat e the e ffect measures .
However, due to the fact that the studies analyzed different
polymorphisms in the same gene, it was decided to perform a
global meta-analysis pooling the same polymorphisms across
the studies, as well as pooling different polymorphisms in the
same gene across the studies. Thus, for the meta-analysis, only
SNPs present in at least two different studies were considered
in the pooled polymorphism and gene results. In addition,
meta-analysis was performed pooled by gene, including re-
sults of individual studies. Moreover, in the analysis, a divi-
sion was made between allele and genotype models, calculat-
ing the estimates for the effect allele and effect heterozygote
and homozygote genotypes, pooling by both polymorphism
and gene. The effect allele and genotypes were compared to
the reference allele and genotype, respectively, in different
analyses. In studies that present more than one categorization
for dental caries, DMF/dmf = 0 vs. DMF/dmf ≥ 1 was chosen.
For the meta-analysis, the results of the adjusted models
(adjusting for ethnicity) were preferably included. In cases
where the adjusted results have not been reported, the unad-
justed estimates were considered or calculated, to be included
in the analysis. In cases where results were only shown by
stratified analysis, we included the group with the highest
number of individuals. The odds ratio (OR) was used to mea-
sure effect size with a 95% confidence interval (CI). The prev-
alence ratio measures were converted to OR using the formula
proposed by Zhang and Yu: PR = odds ratio / 1 − risk0 + risk0
× odds ratio, where “risk0” is the prevalence of disease among
non-exposed individuals [15, 16]. It is important to emphasize
that, in genetic studies, non-genetic factors known to be asso-
ciated with risk of disease can exist in the intermediate path-
ways between the genetic risk marker and disease develop-
ment and, therefore, should be included in the adjusted anal-
ysis to avoid over-adjustment [17]. To address the absence of
the reporting of ethnicity, an investigation was reported of
allele frequencies stratified by populations based on the hu-
man genome (GRCh37.p13).
To avoid inconsistencies with the data analysis, data har-
monization for palindromic SNPs was performed. When the
palindromic SNP was present in two different studies, we only
kept the SNP in the analysis if the study reported the DNA
strand. If this information was missing in the papers, the SNP
was excluded from further analysis. In order to avoid biased
estimates due to linkage disequilibrium (LD) in the gene pool
analysis, a pruning was performed, by LD, for those studies
that analyzed more than one polymorphism in the same gene.
To this end, a pairwise comparison was carried out including
only SNPs which were independent (r
2
< 0.3) from the others.
For the SNPs in LD ≥ 0.3, the analysis included the one with
the lowest p value for the association. When the studies did
not provide estimates of linkage disequilibrium, those re-
trieved from the 1000 Genomes global population as a refer-
ence panel were considered. Thus, when the SNPs included in
the meta-analysis (in gene stratification) were extracted from
Table 1 Search strategy
Search syntax
PubMed #1 (“Dental Decay” OR “Caries, Dental” Or “Decay, Dental” OR “Carious Dentin” OR “Carious Dentins” OR “Dentin, Carious” OR
“Dentins, Carious” OR “Dental White Spot” OR “White Spots, Dental” OR “White Spots” OR “Spot, White” OR “Spots, White” OR
“White Spot” OR “Dental White Spots” OR “White Spot, Dental” OR “Susceptibility, Dental Caries” OR “Caries Susceptibility,
Dental” OR “Caries Resistance, Dental” OR “Resistance, Dental Caries” OR “Dental Caries Resistance”)
#2 (“Polymorphisms, Genetic” OR “Genetic Polymorphisms” OR “Genetic Polymorphism” OR “Polymorphism”
OR “Polymorphisms” OR
“Nucleotide Polymorphism, Single” OR “Nucleotide Polymorphisms, Single” OR “Polymorphisms, Single Nucleotide” OR “Single
Nucleotide Polymorphisms” OR “SNPs” OR “Single Nucleotide Polymorphism”)
Search combination: #1 AND #2
Clin Oral Invest