Our meta-analysis, which examined eight studies encompassing 3232 NHL cases and 4077 controls, investigated the association between SHMT1 C1420T polymorphism and NHL risk. A borderline association was detected, which indicated that this polymorphism may increase NHL risk, although the effect of the SNP was very weak.

SHMT1 is a key enzyme in the folate metabolic pathway and supplies one-carbon molecules to the cycle; this carbon plays an important role in the biosynthesis of purine, thymidylate, and methionine [12], which are essential for DNA synthesis and gene methylation. Therefore, the 1420 C>T polymorphism in SHMT1 creates an imbalance in folate metabolism, which adversely affects DNA synthesis and methylation systems and causes genome instability, eventually leading to overexpression of oncogenes and inactivation of tumor suppressor genes [13, 14]. Additionally, the polymorphism can cause reduced circulating folate levels [23], which not only shunts 5,10-methylenetetrahydrofolate toward DNA synthesis but also results in uracil misincorporation into DNA, eventually leading to double-strand breaks, chromosomal damage, and cancer [31, 32].

We found that SHMT1 C1420T might have a weak effect on NHL risk. There are several possible explanations for this result. First, because only eight studies met our review criteria, the sample size of the meta-analysis was not large enough to detect a specific effect on NHL risk. Second, the cancer risk conferred by the genetic variation is indeed very modest, and the penetrance is very small for the variants. Third, other causal genes, which are implicated in the pathogenesis of cancer, might mask the effect of SHMT1 C1420T polymorphisms by gene–gene interactions and, consequently, modulate cancer susceptibility. In any case, studies with larger sample sizes are warranted to validate our findings.

In the stratifying analysis by ethnicity, we found that SHMT1 C1420T polymorphism might be associated with an increased NHL risk in the mixed ethnic group but not in the Caucasian group, suggesting that there are some differences in genetic information of individuals from different races. In addition, the possibility of misinformation was not ruled out. However, because of the small sample size of the mixed ethnic subgroup, this result remains questionable, and additional studies with larger sample sizes are needed. To date, only one meta-analysis, which was published in 2011 [26], has focused on the association between SHMT1 C1420T and NHL risk. After examining eight studies that encompassed 2884 cases and 4054 controls, Weiner et al. [27] concluded that SHMT1 C1420T had no effect on the risk of NHL, which was inconsistent with the results of our study. Have been examined carefully, we found that the meta-analysis by Weiner et al. [27] included one study by Hishida et al. [33] that we excluded because it provided the data about malignant lymphoma rather than NHL. In addition, we added one more study [29] in our meta-analysis to enlarge the sample size and improve the statistical power. Therefore, compared with the meta-analysis by Weiner et al. [27], our study derives a more precise risk estimate for the association between SHMT1 C1420T polymorphism and NHL risk. Furthermore, we put all the studies together to collectively analyze and then draw a conclusion. Although researchers commonly pool data from different cancer subtypes, it is unclear whether the association of genetic variation with risk of cancer should be calculated this way. We used this approach because many studies have discovered that some sequence variants in specific regions of chromosomes, such as 17p11.2, are associated with risk of specific subtypes of cancer [28, 34]. We speculate that the SNP may be the specific site associated with different NHL subtypes.

Our meta-analysis has several limitations. First, the results of these sensitivity analyses indicated that a single study might change the pooled ORs, which means that our results may have low robustness and should be interpreted cautiously. Second, selection bias resulted from the fact that only studies written in English or Chinese were included in this meta-analysis. Third, the sample size of the included studies was relatively small, especially in the stratified analysis, which may result in limited statistical power. Fourth, significant heterogeneity in the meta-analysis was observed under the recessive model. We did not deem lightly the issue of the random-effects model used to incorporate heterogeneity among studies. Further stratification analysis suggested that ethnicity may be the main source of heterogeneity. Moreover, there is other heterogeneity that cannot be explained. Fifth, because more detailed information was not available in the included studies, possible compounding factors (such as age and sex) could not be obtained for stratification analysis to further evaluate the relationship between SHMT1 polymorphism and NHL risk. Sixth, because all included studies were case–control, this may have caused selection bias, implementation bias, and confounding bias because of the nature of retrospective studies. Finally, in terms of publication bias, the funnel plot’s power is relatively low when fewer than 30 publications are tested for asymmetry. Moreover, most of the data on publication bias are retrospective rather than prospective, including our current analysis. Reporting publication bias from prospective studies is necessary.

In summary, we found in this updated meta-analysis that SHMT1 C1420T polymorphism may be a risk factor for NHL. Additional well-designed studies with larger sample sizes and more information about confounding factors are needed to validate our findings.