|Home | About | Journals | Submit | Contact Us | Français|
Previous studies examining the relationship between the C282Y and H63D HFE mutations and presence of nonalcoholic fatty liver disease (NAFLD) have yielded conflicting results. The goal of this study was to systematically evaluate and summarize data on the association between these two variants and the presence of NAFLD.
The authors searched EMBASE and PUBMED from August 1, 1996 to August 12, 2010. Two investigators independently conducted data abstraction. Ethnic specific weighted prevalence was calculated, and pooled odds ratios were estimated using random effects model.
From 2,542 references, the authors included 16 case-control studies and 14 case-only studies, or 2,610 cases and 7,298 controls. The majority of the studies came from Caucasian populations (2,287 cases and 4,275 controls). The weighted prevalence of HFE mutations in cases was comparable to controls. The meta-analysis was restricted to Caucasians only because the small sample size of non Caucasian participants. The pooled odds ratio for the presence of any HFE genetic variant in cases was 1.03 (95% CI: 0.90, 1.17; I2: 65.8%, 95% CI: 38.5, 81.0). The presence of other genotypes, and secondary analyses yielded similar non significant findings.
Our systematic review does not support an association between the HFE genetic variants and the presence of NAFLD.
Nonalcoholic fatty liver disease (NAFLD), the most common chronic liver disease , represents a wide spectrum of disease characterized by the presence hepatic steatosis in the absence of significant alcohol consumption or other causes of liver disease [2, 3] The pathogenic processes leading to steatosis, steatohepatitis and fibrosis are multifactorial and involve both environmental and genetic factors [4, 5]. Iron overload has been reported in NAFLD and could contribute to its pathogenesis and progression. Iron overload influences lipid [6, 7] and glucose metabolism , and thus may lead to insulin resistance, a key factor in the pathogenesis of NAFLD. Furthermore, iron excess may also contribute to progression of NAFLD thru oxidative stress and fibrogenesis within the liver [9, 10].
Feder et al. identified in 1996 the cysteine-to-tyrosine substitution at the amino acid 282 (C282Y), and the histidine-to-aspartate change at the amino acid 63 (H63D) variants of the HFE gene (OMIM 235200) . These variants are responsible for the majority of cases of hereditary hemochromatosis, and are very common in Caucasians (6.2% and 15.1%, respectively) . Compared to other genotypes, C282Y homozygotes are at greatest risk of iron overload, followed by compound heterozygotes with one C282Y mutant allele and one H63D mutant allele; on the contrary, simple heterozygotes for both C282Y and H63D may also develop increased serum iron markers, but to a lesser extent [13–16]. Since NAFLD and HFE mutations are common, many authors studied the association between these two conditions and found conflicting results. Inadequate sample size, referral and ascertainment biases, different ethnic background, lack of mediators’ adjustment, or publication bias could explain such inconsistencies. A wide scope meta-analysis published suggested a positive association for some genotypes with NAFLD ; in this paper, however, we performed a meta-analyses focused in NAFLD with updated literature search.
The main goal of this study was to systematically review the available evidence examining the association between the common HFE genetic variants (C282Y and H63D) and NAFLD. We specifically studied whether the HFE mutations were enriched in NAFLD patients compared to non diseased controls; we also studied the association between C282Y homozygotes and C282Y/H63D because these are the most iron loading genotypes. We hypothesized that the HFE genetic variants would be more frequent in NAFLD patients compared to controls. A positive association would suggest that HFE could identify patients at risk for NAFLD and, that iron depletion by phlebotomy could be explored as a safe and easy therapy for NAFLD-HFE (+) patients.
We aimed to identify all studies that assessed the association of the C282Y and H63D genetic variants and NAFLD using a previously defined protocol . We searched PUBMED and EMBASE for the period between August 1st, 1996 (when HFE was firstly described)  and August 12th, 2010 using free text and subheadings with explosion (MeSH and EMTREE terms) and without language restriction (Appendix A and Figure S1). In addition, we reviewed the reference lists of relevant original papers and review articles. We excluded papers without original data, animal, or in vitro studies. We also excluded those papers whose inclusion criteria required carrier status for HFE mutations . Lastly, we also excluded studies that reported only extreme phenotypes (hepatocellularcarcinoma and cirrhosis) since their inclusion may introduce phenotypic heterogeneity.
Given that some studies in Caucasian populations did not have a control group, we performed secondary analyses imputing a control group from the non Hispanic Whites (NHW) of the Third National Health and Nutrition Examination Survey (NHANES III) to enrich the number of participants .
Two investigators (R.H. and E.Y) independently reviewed the search results to determine study eligibility and to perform data abstraction. Discrepancies were resolved by group consensus. For each selected publication we abstracted key study characteristics and genotype frequencies. When the genotypes of interest were not directly shown, we tried first to calculate them from the published data [20, 21], or, if not possible, we contacted the authors. For the quality of the reporting, we used a modified STREGA checklist (Appendix B) , and MOOSE recommendations .
Since both C282Y and H63D are functional variants [24, 25], confounding by population stratification may be less of a concern in this study. That is, both mutations have the same effect across different ethnicities. Since the HFE genotype distribution [26, 27] and NAFLD  prevalence varies between ethnicities, we made the a priori decision to stratify our analyses by ethnicity. Since the majority of the papers did not explicitly report the ethnicity/race of the participants, we decided to consider ‘Caucasian’ those studies which reported a majority of Caucasians (e.g. least 70%), or that were conducted in countries whose majority is mainly Caucasian (i.e. Europe, Australia and North America).
For those studies that included a control group, we computed the odds ratios (OR) and 95% confidence intervals for NAFLD comparing the presence of any HFE mutation (C282Y/C282Y, C282/H63D, C282Y/WT, H63D/H63D, H63D/WT) to non diseased controls. We also obtained pooled OR for all possible genotypes, and created different subgroups such as those genotypes traditionally associated with iron overload (C282Y/C282Y and C282Y/H63D), or mixed heterozygotes (C282Y/H63D, C282Y/WT, and H63D/WT).
In order to combine those studies without control group, we imputed the genotype frequencies from the non Hispanic Whites population in a representative sample of the U.S. population . We used the U.S. weighted frequencies reported by Steinberg et al. (C282Y/C282Y: 0.30%; C282Y/H63D: 2.35%; H63D/H63D: 2.15%; C282Y/WT: 9.54%; H63D/WT, 23.55%; and, WT/WT: 62.10%. 
We considered the reference group those individuals wild type at both loci (WT/WT). We used the Cochran’s Q-test with a significance level of p≤0.10 to qualitatively assess statistical heterogeneity , and the I2 statistic with the 95% confidence intervals to quantify the proportion of total variability in point estimates that can be attributed to heterogeneity . To account for possible statistical heterogeneity, we adopted the inverse variance method with random effects model to calculate the ethnic specific pooled odds ratios [30–32]. If statistical heterogeneity was present, we analyzed the impact on the interpretation of pooled OR by stratifying on clinically relevant characteristics. We decided to provide a pooled estimate when, despite statistical heterogeneity, the inference remained unchanged. We assessed also the presence of publication bias by the Harbord regression test for funnel-plot asymmetry . All analyses were done using the updated commands metan, metabias, and metainf available in STATA 9.2 (College Station, Tx). A portion of these analyses were presented at the 60th American Association for the Study of Liver Disease Meeting .
We identified 2,542 unique references (2,272 in EMBASE; 1,602 in MEDLINE), of which 2,359 did not meet our inclusion criteria. We excluded 153 studies due to lack of data on the HFE genotypes in NAFLD participants; whose liver disease was not defined as NAFLD; or whose participants were reported included in separate reports. A total of 16 case-control studies [20, 21, 35–49] (Table 1) and 14 ‘case-only’ studies [50–63] (Table 2), or 2,610 cases and 7,298 controls were included for the current review.
Overall, twenty-two studies were originated in Europe, North America, or Australia [20, 21, 35, 37–40, 42–45, 47, 48, 50, 51, 54–60, 63], two studies in Brazil [49, 52] and Japan [61, 62], respectively and, one study each in Korea , Taiwan , India and Turkey . The quality of the reporting varied across the papers. Whereas the indication for the liver study, the NAFLD definition, and the outcome ascertainment among cases were usually reported, only nine studies reported the ethnicity of the participants [20, 36–38, 47, 49, 52, 58, 62, 63], three reported some quality control in the DNA collection [36, 37, 40], and three excluded C282Y homozygotes from the study [36, 42, 48, 58]. (Appendix C). The most common reason for investigation of possible NAFLD was elevation in liver enzymes and/or an “echogenic” liver on an ultrasound examination. The method for ascertainment of NAFLD cases was pathology in most studies, and most cases in such studies were nonalcoholic steatohepatitis (NASH), rather than simple steatosis. For definition of controls, seven used clinical and imaging criteria to rule out NAFLD [35, 36, 43, 44, 47–49], and others used blood donors [20, 39, 47], newborns [21, 45, 46] or did not describe the methodology [37, 38, 41, 42]. Specifically for the case-control studies, the majority of the participants had presumed Caucasian ethnicity (1,727 cases and 4,275 controls), three had iron overload as the criteria for inclusion [42, 44, 48], six used newborns or blood donors as controls [20, 21, 39, 45–47], and four studies used only ultrasound to define NAFLD [36, 40, 41, 44].
The internally weighted prevalence for the HFE genotypes in people with NAFLD is depicted in Figure S2. Among Caucasians, the prevalence of C282Y and H63D homozygotes was 1.01% (95% confidence interval, CI: 0.60, 1.41%) and 2.01% (95% CI: 1.44, 2.59%), respectively. The prevalence of compound heterozygotes, C282Y heterozygotes, and H63D heterozygotes were 1.66% (95% CI: 1.14, 2.19%), 10.76% (95% CI: 9.49%, 12.03%), and 16.70% (95% CI: 15.17, 18.23%), respectively, which are comparable to the control group and non-Hispanic White U.S. population . In contrast, no C282Y mutation was present in non-Caucasian populations, which had H63D heterozygotes as the most common genotype after WT/WT.
The odds ratio for NAFLD among Caucasians was not significantly increased for those carrying HFE mutations (C282Y or H63D) compared to controls (OR: 1.03, 95% CI: 0.90, 1.17). Among non-Caucasians, there was a significant association (OR: 1.64; 95% CI: 1.20, 2.24), Figure. There was stronger evidence for statistical heterogeneity in the odds ratio between Caucasians (I2: 65.8%, 95% CI: 38.5, 81.0) to non-Caucasians (I2: 0%; 95% CI: 0, 85.1). Given the small number of studies in non Caucasian population, we restricted the following analyses among Caucasians only. We addressed statistical heterogeneity by studying different clinically relevant characteristics as previously described. Our inference of no association between the presence of any HFE mutation and NAFLD remained unchanged. Considering that previous authors encountered association with a combination of heterozygotes  we conducted similar analyses and found no association. Similar results were also found with a category that included both C282Y/C282Y and C282Y/H63D, the most penetrant (iron loading) genotypes. All the previous analyses were also stratified by clinical characteristics (Supplementary Table) and reached similar conclusions without statistical evidence for publication bias as shown by Harbord tests (Figure S3).
Finally, in order to include the genotypes of those 14 studies without control group, we imputed the genotype frequencies using the non-Hispanic white population and found that the presence of any HFE mutation was not associated with NAFLD (OR: 1.00, 95% CI: 0.87, 1.14),. When considering different genotypes, only the presence of C282Y/C282Y was associated with NAFLD (OR: 6.50, 95% CI: 2.30, 18.35), and other categories including C282Y homozygotes (Figure S4).
Our current systematic review and meta-analysis of 13 case-control studies did not find a significant association between the presence of HFE mutations and NAFLD in Caucasians. Specifically, we showed that the presence of iron overloading genotypes (C282Y/C282Y, C282Y/H63D) did not increase the risk of having NAFLD compared to controls. When we imputed controls from NHANES, we found similar inferences except from the presence of homozygosity. This may be explained by the presence of oversampling of cases with C282Y homozygosity among the case-only studies rather than a real association. In non Caucasian populations, our results based on three case-control studies suggest an excess of H63D mutants in NAFLD cases compared to controls.
In contrast to Ellervik et al , our analyses showed no association due, we believe, to three key factors. First, applying the criteria they described, two studies should not have been excluded [45, 49], and, among the studies included, Ellervik et al. did not abstract the studies with the greatest number of cases [37, 38, 47] in the meta-analysis. Secondly, weights used for meta-analysis appear to be inconsistent from the confidence intervals estimated by Ellervik and colleagues (Ellervik et al., Webfigures 6 and 9). In other words, using the abstracted study-specific odds ratios and 99% confidence intervals (Webfigures 6 and 9) we found incorrect weights and, consequently, inaccurate pooled odds ratios for the C282Y/C282Y genotypes and mixed heterozygotes, but not for the H63D/H63D genotype. Finally, two of the studies abstracted by Ellervik et al. did not genotype H63D in all individuals and this fact was not taken into account in the prior meta-analysis [20, 21]. An individual heterozygous for one mutation may also be heterozygous for another mutation (compound heterozygotes); therefore, it is possible that, if the reference group with genotype WT/WT may contain three other types of people, namely those with H63D/H63D, C282/H63D and H63D/WT. As a consequence, Ellervik et al. used incorrect study-specific odds ratios because they did not take into account this difference in the genotyping frequencies, and therefore, overestimated the pooled estimates for the C282Y/C282Y (OREllervik:10.42, 99% CI: 2.05, 52.99; ORcorrected: 5.87, 99% CI: 1.47, 23.48) and for the mixed heterozygotes genotypes (OREllervik: 3.30, 99% CI: 1.40,7.70; ORcorrected: 1.47, 99% CI: 1.01, 2.12). In contrast we found similar estimates for the H63D homozygotes (OREllervik: 1.30, 99% CI: 0.49, 3.43; ORcorrected:1.25, 99% CI: 0.50, 3.10) (Figure S5). When we applied those corrections and replicated the Ellervik et al. meta-analysis focusing on their search period, we did not find any association between HFE and NAFLD (data not shown).
The lack of power, different ethnic variation, information and selection biases and the presence of confounders have been frequently quoted to explain the inconsistency across HFE-related studies [20, 58, 64]. A meta-analysis is a possible solution to overcoming the issue of power. In our study of 1,727 NAFLD Caucasian cases and 4,275 controls, we had 88% power to detect an OR of 1.20 assuming a prevalence of any HFE mutation in the reference population of 36.9% (estimated from our findings) and a two-sided alpha error of 0.05 . Therefore, it is likely that even if HFE variants had an effect on the presence of NAFLD, the effect size would be relatively small. We also showed that information/selection biases did not substantially influence our results. Our results showed a trend to increased odds in non Caucasian populations, however, this finding deserves more studies since the overall estimate is driven by largest of the three studies[36, 46, 53].
The current analysis has several limitations. First of all, we weren’t able to address whether the HFE genetic variants acted as a disease-modifying gene in NASH (e.g. compare genotypes in simple steatosis only to those in NASH) or the impact of HFE genetic variants on iron biomarkers in NAFLD patients. The published data were insufficient to address such question; and remains unsolved [58, 66]. Our definition of ethnicity was somehow arbitrary but, we believe, was consistent with the reported literature. It also reflects the unusual reporting of ethnicity across studies, a key variable in genetic studies. Third, the extrapolation of these results in the general population is somehow difficult because the majority of the case-control studies came from the hospital setting (where cases were defined using histology). Finally, when we examined statistical heterogeneity across different clinically relevant characteristics as previously described, our inference of no association between the presence of any HFE mutation and NAFLD remained unchanged. We acknowledged that we did not have enough characteristics of each study to correct for the presence of statistical heterogeneity. Only cooperation between researchers and individual data meta-analysis could address this question using meta-regression or genome wide association studies.
In contrast, our meta-analysis has several strengths. Not only we systemically reviewed the literature but also we obtained primary data from authors if needed. Compared to Ellervik, our current initiative focus and expands our knowledge on the association between the presence of HFE genetic variants and NAFLD.
In conclusion, we have found no association between the HFE genotype and NAFLD in Caucasians and non-Caucasians. We believe the HFE may have none or, at maximum, a marginal role in the development of NAFLD. Future studies should focus on the role of HFE in the progression of NAFLD and its association with iron biomarkers (peripheral and hepatic).
The authors are deeply grateful to the following authors that provided unpublished data: Drs. Yves Deugnier and Michel Mendler; Dr. Luca Valenti; Dr. Paola Loria; Dr. Stergios Kechagias; Drs. Vincenzo Baldo and Annarosa Floreani, Dr. Sergio Neri; Dr. Laurence Vénat-Bouvet. The authors also thanks the following authors for the correspondence: Dr. Bruce Bacon, Dr. SH Jeong; Dr. Lawrence C. Nichols; Dr. Paul C. Adams; Dr. Zeynel Mungan, Barbara Banner; Dr. Jose Antonio Solis Herruzo; and, Dr. Ajay Duseja
This work was supported by the following grants: the American Diabetes Association, Mentor-Based Fellowship Program 7-07-MN-08 (RH, FLB);NIDDK Diabetes Research and Training Center, P60 DK079637 (FLB) and NIDDK Patient Oriented Research in Type 2 Diabetes grant, K24-DK62222 (FLB); NIH grant DK-02957 and the Virginia Mason Liver Center of Excellence Fund (KVK); and K01DK067207 (WHLK)
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.