In this first meta-analysis of the association between the MMP1 -1607 1G>2G polymorphism and risk of HNC, we found significant associations in the overall comparison using the recessive and allelic contrast models. Individuals with the 2G/2G genotype or 2G allele carriers could have an increased risk of HNC. Moreover, in the stratified analyses by several variables, including ethnicity, source of the controls, and tumor site, significant associations were observed in the Asian population, European population, population-based control subgroups, oral cavity cancer, and nasopharyngeal cancer. Although our analysis had relatively small sample size, with current sample size, we, however, had power to detect a reasonable degree of association. These results suggest that the MMP1 -1607 1G>2G polymorphism might modulate genetic susceptibility to HNC.
MMP1, a major member of the MMPs family, has been implicated in the development of a variety of cancers because of its ability to degrade ECM 
. The expression level of the MMP1
gene can increase in various tumors, which has been associated with a poor prognosis in some types of cancers 
. Moreover, the promoter region of the MMP1
gene can influence its expression. Rutter et al. first described the polymorphism at -1607 in the MMP1
. It has been demonstrated that the MMP1
-1607 1G>2G polymorphism is associated with increased transcription of the MMP1
gene which is attributed to its 2G allele creating a core-binding site for the Ets transcription factor family, resulting in a higher expression level of MMP1
In this meta-analysis, we found that individuals with the 2G/2G genotype had a higher risk of developing HNC under a recessive model, but no association was observed under a dominant model, which implies that homozygous 2G may have a stronger effect on an individual’s phenotype than heterozygous 2G, and thus, 2G/2G genotype carriers may be more susceptible to the development of HNC than 1G/2G or 1G/1G genotype carriers. Similarly, we also found that under the allelic contrast model 2G allele carriers had a higher risk of HNC than 1G allele carriers. This finding suggests that the 2G allele may increase susceptibility to HNC because of its association with increased transcription of the MMP1 gene. However, this hypothesis needs to be tested in future studies.
A few of the source studies also reported results linking the 2G allele to an increased risk of HNC. O-charoenrat et al 
found that cell lines with 2G/2G genotype expressed a higher level of MMP1
mRNA than other genotypes and individuals with the 2G/2G genotype had a higher risk of HNC, suggesting that the MMP1
2G allele may be a risk factor that could increase susceptibility to HNC. Cao et al. investigated the role of the MMP1
-1607 1G>2G polymorphism in oral squamous cell carcinoma (OSCC) and reported that the 2G allele increased significantly in OSCC patients when compared to controls, indicating that the MMP1
-1607 1G>2G polymorphism may be associated with risk of OSCC in a Chinese population 
. Similarly, Nishizawa et al. explored the association between MMP1
-1607 1G>2G and risk of OSCC in a Japanese population and found that the frequency of 2G alleles was significantly higher than that of 1G allele in OSCC patients 
. They concluded that the MMP1
2G allele might play a crucial role in the early onset of OSCC. However, Zhou et al. reported that no significant association between the MMP1
-1607 1G>2G polymorphism and risk of HNC was found in two different populations 
. These inconsistent results may be attributed to differences in genetic backgrounds, environmental factors, and other factors, such as small sample size or inadequate adjustment for confounding factors.
Interestingly, our subgroup analysis by ethnicity showed that the MMP1
-1607 1G>2G polymorphism played different roles in Asian and European populations. In the European population, it was significantly associated with reduced risk in all three genetic models in European population, whereas in Asian population, it was significantly associated with increased risk. For example, Zinzindohoue et al. examined the association of the MMP1
-1607 1G>2G polymorphism with risk of HNC in a case-control study in a European population 
. They found that 2G allele frequency was significantly lower in cases than in controls, and individuals with the homozygous 2G/2G genotype were at lower risk of HNC than those with the 1G/1G genotype. Similarly, Vairaktaris et al. found that the MMP1
-1607 1G>2G polymorphism was associated with a decreased risk of oral cancer in 2G allele carriers in a European population 
. In contrast, in Asian populations, most studies found that the MMP1
-1607 1G>2G polymorphism was associated with an increased risk of HNC in patients with the 2G/2G genotype or 2G allele carriers 
. These conflicting results may be due to the different genetic backgrounds in these populations, subsequently leading to different genetic susceptibility to the same disease. Moreover, HNC is a disease caused by multiple genetic and environmental factors, and possibly gene–gene and gene–environment interactions. Additionally, other factors such as linkage disequilibrium (LD) may also contribute to this discrepancy 
. However, because of the limited number of studies in European population and relatively small sample sizes, these results should be interpreted with caution. Further study with larger sample sizes is warranted in different populations.
Heterogeneity is a major problem when interpreting the results of meta-analyses. In this study, significant heterogeneity was detected in overall comparisons using all three genetic models. Ethnicity was an important reason for this heterogeneity. Individuals from different ethnicities may have diverse genetic backgrounds and environmental factors, and consequently, the same polymorphism may play different roles in different populations. Therefore, when we performed stratified analysis by ethnicity, the heterogeneity disappeared in the European population and decreased significantly in the Asian population. Furthermore, the source of the controls was another factor that contributed to heterogeneity. The MMP1 genotype distributions in population-based controls may be similar to normal, and thus, population-based controls could be more reliable than hospital-based controls. This might partially explain why the results from the stratified analysis by the source of the controls were different between the two subgroups. In addition, another reason for the heterogeneity between studies was the tumor site. In the stratified analysis by tumor site, significant associations were found for oral cavity cancer and nasopharyngeal cancer, but not for either pharyngeal (oropharynx/hypopharynx) cancer or laryngeal cancer. Although HNC includes tumors from different sites, risk factors for these cancers are different. For example, oral cavity and laryngeal cancers are majorly associated with tobacco use and alcohol consumption, while oropharyngeal and nasopharyngeal cancers are principally related to viral infection, such as human papillomavirus (HPV) and Epstein-Barr virus (EBV). Thus, further studies with larger sample size and different tumor sites are warranted.
The present study has some limitations. First, the study number was limited and the total sample size was relatively small; thus, our estimates of association might have occurred by chance. Second, significant heterogeneity was detected in our study, and thus, the results must be interpreted with caution. However, heterogeneity disappeared in some subgroups when stratified analysis was performed. Therefore, the results from the subgroup analyses may be more meaningful, as the polymorphism may play different roles in diverse subgroups. Third, further subgroup stratification based on other risk factors such as alcohol consumption, tobacco smoking and HPV status could not be performed because of the limited data 
. Fourth, our meta-analysis was based on unadjusted estimates because only 3 original studies provided adjusted estimates, and the adjusted covariates varied among these studies. A more comprehensive analysis should be conducted if detailed information such as environmental factors and lifestyles are available. Finally, we could not conduct a meta-analysis using linkage disequilibrium, as few studies performed haplotypic analysis.
In conclusion, our meta-analysis suggests that the MMP1 -1607 1G>2G polymorphism is associated with HNC risk. Moreover, subgroup analysis based on ethnicity indicates that it may play different roles in Asian and European populations. However, due to the limited study numbers and relatively small sample sizes, our results should be validated in future studies with larger sample sizes and in different ethnic populations.