Reply

Abstract

We appreciate Dr Braillon’s interest in our recent work.1Yang W. et al.Clin Gastroenterol Hepatol. 2020; 18: 2775-2783Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar Although modification of dietary habits is a promising approach for cancer prevention, we are at an earlier discovery phase for research on diet and hepatocellular carcinoma (HCC) than, for example, colorectal cancer (CRC).2Lee D.H. et al.Am J Public Health. 2016; 106: 1599-1607Crossref PubMed Scopus (41) Google Scholar By using the data from the Nurses’ Health Study and the Health Professionals Follow-up Study, we published a number of earlier prospective studies on a number of risk factors for CRC in the 1990s.2Lee D.H. et al.Am J Public Health. 2016; 106: 1599-1607Crossref PubMed Scopus (41) Google Scholar We reported that alcohol, smoking, red and processed meat, and obesity were associated with an increased risk, whereas folate, calcium, vitamin D, and physical activity were associated with a decreased risk, of CRC. These studies may have been interpreted as p-hacking at the time, but all of these findings were confirmed by subsequent studies as summarized by the World Cancer Research Fund/American Institute for Cancer Research report and contributed to the field of CRC.3World Cancer Research FundContinuous Update Project Report. Food, Nutrition, Physical Activity, and the Prevention of Colorectal Cancer. American Institute for Cancer Research, Washington, DC2011Google Scholar Indeed, our study population consisted largely of white, educated, health professionals. The relative homogeneity was an advantage because the internal validity is more crucial than external validity at this stage. In addition, in a prospective study, however, selection of participants typically does not affect estimates of measures of associations (such as rate ratios) because outcomes (unobserved at the time of selection) are not correlated with exposures for reasons other than causality and confounding, which is addressed in the analysis. Furthermore, the HCC annual incidence in our study cohorts was 2 per 100,000 in women and 5 per 100,000 in men, which corresponds to population-based estimates for the time period of our study,4Yang W. et al.JAMA Oncol. 2019; 5: 1509-1510Crossref Scopus (1) Google Scholar suggesting that our findings remain quite relevant for a substantial portion of the US population. The Nurses’ Health Study and the Health Professionals Follow-up Study cohorts have collected detailed and validated information on potential confounding factors. In our study,1Yang W. et al.Clin Gastroenterol Hepatol. 2020; 18: 2775-2783Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar we adjusted for a large number of covariates including sex, race, smoking, type 2 diabetes, body mass index, and other lifestyle factors such as physical activity and dietary factors (alcohol, coffee, and total calorie intake). We also controlled for the overall impact of diet (ie, the 2010 Alternative Healthy Eating Index score) in a sensitivity analysis and yielded similar results. Although we cannot adjust for hepatitis B or C virus, these hepatitis virus infections were not correlated with fat intake, arguing against substantial confounding by hepatitis. We found essentially the same results after further adjustment for smoking with pack-years. We also conducted a priori exploratory subgroup analyses and did not find any statistically significant interactions between fat intake and other risk factors.1Yang W. et al.Clin Gastroenterol Hepatol. 2020; 18: 2775-2783Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar Statistical significance clearly is useful when interpreting study results, although we do not rely purely on it to make causal inference. Although the magnitude of the hazard ratio alone cannot be used to explain the magnitude of the relative risk in some situations,5Sutradhar R. et al.Ann Epidemiol. 2018; 28: 54-57Crossref PubMed Scopus (46) Google Scholar the hazard ratio is still a valuable estimate for understanding time-to-event data, and there typically is no need to avoid its use.5Sutradhar R. et al.Ann Epidemiol. 2018; 28: 54-57Crossref PubMed Scopus (46) Google Scholar Although the current epidemiologic studies of fat and HCC are limited, we found that intake of vegetable and polyunsaturated fat is associated with a lower risk and these findings are biologically plausible. In randomized intervention studies, saturated fat dramatically increased hepatic fat relative to polyunsaturated fat.6Rosqvist F. et al.Diabetes. 2014; 63: 2356-2368Crossref PubMed Scopus (234) Google Scholar It is premature to report population-attributable fractions because the association has not been established as causal. Diets and Hepatocellular Carcinoma: The Emperor's New ClothesClinical Gastroenterology and HepatologyVol. 19Issue 2PreviewThe conclusion by Yang et al1 that “replacing animal or dairy fats with vegetable fats, or replacing saturated fats with monounsaturated or polyunsaturated fats, was associated with reduced risk of hepatocellular carcinoma (HCC)” deserves robust comment. Full-Text PDF