Reply

Potential conflict of interest: Dr. Younossi consults for Gilead, Bristol‐Myers Squibb, AbbVie, Intercept, and GlaxoSmithKline. We appreciate the significant interest that has been raised by our recent meta‐analysis.1 In their letter, Lonardo et al. raised some important issues regarding the association of nonalcoholic fatty liver disease (NAFLD) with metabolic syndrome, the use of the National Health and Nutrition Examination Survey (NHANES) data set for this meta‐analysis, and cardiovascular (CV) mortality in patients with NAFLD. On the first point, we agree with the authors that NAFLD may be a precursor of metabolic syndrome rather than its hepatic manifestation. Which condition came first or which one is the precursor or the consequence of the other is difficult to ascertain. It is more plausible that these conditions are part of the spectrum of overlapping metabolic pathways leading to the hepatic phenotype of NAFLD and extrahepatic metabolic diseases. The use of NHANES for this analysis requires some clarification. In this analysis, we intentionally included multiple studies which had used the NHANES data set because this approach may have given us a better estimate of the prevalence rates more representative of the general population. Given that different studies used different designs, we did not want to arbitrarily choose one study over another. We have subsequently rerun our analysis by including only one of the NHANES studies. In fact, if we limit the inclusion to only one NHANES, the prevalence of NAFLD in North America (primarily from the United States) did not change (23.64%). A final point that was raised by the authors is related to CV mortality in NAFLD patients. In fact, our analysis did show that CV events are the most common cause of death in NAFLD patients and that the incidence of CV mortality was higher than liver mortality.1 However, the incidence rate ratio initially did not show increased CV mortality in NAFLD patients compared to the controls. We believe this is related to the differences in the definitions of NAFLD and the modalities that were used to establish the diagnosis of NAFLD. If we separate these studies based on the diagnostic modalities (radiologic versus blood tests), CV mortality was indeed higher in those with ultrasound NAFLD (incidence rate ratio = 1.37, range 1.23‐1.54). On the other hand, CV mortality in NAFLD patients diagnosed by liver enzymes alone was not increased.1 One explanation for this finding may be the high rate of CV mortality in the control groups, which is consistent with heart disease being the most common cause of death in the United States. Additionally, these findings could suggest that the presence of fatty liver by ultrasound has a closer association with CV disease than the presumed diagnosis of NAFLD by elevated liver enzymes. In our article, this increase in the risk of CV mortality of NAFLD by ultrasound is clearly stated in the seventh paragraph of the discussion.1 In the second letter, by Roereck et al., the use of data from NHANES was again questioned. As we indicated above, at the onset of the analysis, we had to decide whether to include one of the NHANES studies (arbitrarily) or to include all of them in the hope of getting closer to the true prevalence of NAFLD in the general US population. As reported, this prevalence rate was estimated to be 24.13% (19.73%‐29.15%).1 In response to these comments, we reanalyzed the data by including only one report from NHANES that had the largest sample size and used ultrasound to establish the diagnosis of fatty liver which were the most recently published data. The recalculation of the prevalence for North America (primarily from the United States) was 23.64%, which is almost identical to the rate initially reported. We also consulted with other methodologists about the use of data from different studies based on the NHANES data set. In fact, the authors may not fully understand the design of NHANES as a data set. Just because multiple studies were sampled from the same population data set (NHANES) does not mean the samples were exactly the same. In fact, the more samples used, the more representative the results are for the NHANES subjects enrolled from the US population. Another issue raised the possibility of not including some studies from Japan and Asia. As indicated in the article, we started the study with 728 articles. It is possible that some studies were not included because they were not fully published at the time of the search or did not meet our selection criteria. In this context, we would hardly consider the immense review of all the published data summarized in the article and supporting tables as incomplete. We completely disagree with Roereck et al. about the validity of our results. Not only were our data collection and analyses very in‐depth and rigorous, but the results of our analysis show face validity. We do, however, encourage Roereck et al. to use their methodologic expertise and perform another similar meta‐analysis or to at least generate epidemiologic data about NAFLD from Canada which is currently quite sparse. Finally, we appreciate the comments by Naderian et al. pointing out the high prevalence of NAFLD in Iran. In fact, if we take the average of the prevalence rates provided by them for different regions in Iran, these rates are close and within the range of the rate reported by us for the Middle East (31.79%, range 13.48‐58.23). Additionally, we were unable to locate in PubMed the study cited in their table with the most recent prevalence rates from Iran. Nevertheless, the authors do bring up a very important issue related to the increasing prevalence of obesity, metabolic syndrome, and NAFLD in the Middle East. In this context, we encourage the authors, as well as other investigators, from this region to continue to publish in this area and raise awareness about this important liver disease in the Middle East. We express our appreciation to all the authors for pointing out issues that needed clarification in our study. In fact, we believe that our meta‐analysis provides crucial data to estimate the clinical burden of NAFLD. It also brings out important issues related to the selection of study populations and diagnostic modalities for NAFLD. We observed that in different epidemiologic studies authors have used liver enzymes, noninvasive biomarkers, radiologic modalities, or liver biopsy to define NAFLD. Each one of these diagnostic modalities has its own performance characteristics that can affect a study's conclusions. Furthermore, while some authors used population‐based data, others have reported from selected populations (diabetics, referral center cohorts, etc.). Conclusions from these different study designs are generalized to the entire population of NAFLD. Finally, there is a paucity of robust epidemiologic data from certain parts of the world. In this context, meta‐analysis or systematic review can provide an estimate of disease burden for NAFLD. It is time to have a global effort to carefully estimate the true prevalence of NAFLD in different regions of the world, to agree on very stringent criteria for diagnosing NAFLD and nonalcoholic steatohepatitis, and to recommend what outcomes are clinically important (e.g., progression to stage 2 fibrosis, liver‐related mortality, CV mortality, and/or overall mortality).2 In summary, this very large and in‐depth analysis clearly shows that NAFLD is the most common cause of liver disease worldwide. It also shows that diagnosis of NAFLD increases the risk for adverse outcomes. It is time to develop a consensus about diagnostic criteria, study population, appropriate outcomes for NAFLD, as well as a multidisciplinary and multifaceted approach to studying NAFLD and its impact on mortality and morbidity worldwide.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.