Nonalcoholic fatty liver disease (NAFLD) has become the most common cause of chronic liver disease, affecting roughly 80 million individuals in the United States (
). The rapid increase in end-stage liver disease and hepatocellular carcinoma related to NAFLD and its severe phenotype, nonalcoholic steatohepatitis, has resulted in a considerable economic burden and stress on healthcare systems worldwide ( 1,2 ). Cardiovascular diseases (CVD) and malignant neoplasms have been identified as the commonest causes of death in NAFLD ( 3,4 ). Thereby, identifying modifiable lifestyle factors for NAFLD and its leading causes of mortality could potentially reduce its population burden. Physical activity (PA) and diet quality (DQ) are modifiable risk factors shown to be strongly associated with NAFLD and its severity ( 5–7 ). Observational and interventional studies have reported that increased PA, including different modalities, and a better DQ are associated with a lower risk of NAFLD, nonalcoholic steatohepatitis, and advanced fibrosis ( 8–16 ). 8–15
In several large prospective cohort studies, moderate to vigorous intensity PA (MVPA) and higher DQ as assessed by the Healthy Eating Index (HEI), Alternate Healthy Eating Index (AHEI), and Dietary Approaches to Stop Hypertension score have been associated with a lower risk of chronic diseases, and all-cause, CVD-related, and cancer-related mortality (
Most of the evidence on the association between PA and mortality has been derived from studies using self-reported PA levels, which are prone to recall and response bias and may overestimate or underestimate actual PA energy expenditure and rates of inactivity (
). The objective measurement of PA using accelerometers is believed to overcome several of the limitations associated with self-report measures. Accelerometers may capture data on movement and intensity every minute, potentially allowing estimation of the duration and intensity of activity with greater precision than self-reports ( 23 ). Hip-worn accelerometers may reflect whole-body movement and thus energy expenditure, but poor wear time compliance, which ultimately leads to less accurate estimates of PA and inactivity, and thus has seen an increased use of wrist-worn accelerometers to assess habitual PA ( 24 ). 25
In 2011–2014, the National Health and Nutrition Examination Survey (NHANES) deployed ActiGraph GT3X+ wrist-worn accelerometers to measure movement behaviors. Advantages to this placement include a lower participant burden that leads to higher compliance, a more secure attachment of the device, and 24-hour, including sleeping time, data. Acceleration measurements obtained on the x-, y-, or z-axes with ActiGraph GT3X+ wrist-worn accelerometer are summarized in Monitor-Independent Movement Summary (MIMS) units, which is generated by a nonproprietary device–independent universal algorithm that will allow comparisons across various studies and designs (
). MIMS is used to quantify the total continuous movement performed throughout a 24-hour period, with larger MIMS units (i.e., minutes) indicating higher amounts of movement. Equivalent to other traditional PA metrics (i.e., daily steps or activity counts), PA volume can be expressed as MIMS/min per day, which corresponds with the total MIMS units accumulated per day across valid days of assessment ( 26 ). Conventionally, accelerometer-measured intensity has been expressed by the number of minutes per day or week spent in MVPA, using manufacturer-specific cut points, which correspond to 3 METs ( 27 ). Until recently, an equivalent intensity-based expression of MIMS units has been lacking. Recently, the concept of peak 30-minute cadence (i.e., the average of the 30 highest cadence [steps/min] values within a day) has been introduced to represent an individual's “best effort” ( 28 ). Peak 30-minute cadence is believed to capture intensity and persistence of movement behavior within a day and its regularity across valid days ( 29 ) and has been used as a proxy of PA intensity in recent studies ( 30,31 ). A strong correlation has been observed between MIMS/min and higher PA intensity ( 32,33 ). Thus, akin to the peak 30-minute cadence, minute level of PA in NHANES data could be used to compute peak 30-minute MIMS as an indicator of higher-intensity epochs over the monitoring period. The examination of daily accumulated MIMS (volume) and peak-30 MIMS (intensity) allows a comprehensive evaluation of PA and its relationship with mortality. 26
The HEI is a measure of DQ used to assess how well a set of foods aligns with key recommendations of the Dietary Guidelines for Americans (DGA) (2015–2020) (
). The DGA recommends a healthy diet encouraging eating a variety of vegetables, fruits, whole grains, low-fat dairy, and protein foods within an appropriate calorie level while limiting consumption of saturated fats and trans fats, added sugars, and sodium ( 34,35 ). The HEI-2015 incorporates 13 components assessing foods' compliance with the DGA 2015–2020 (see more details on the Supplementary Digital Content, see Supplementary Material, 35 ) ( https://links.lww.com/AJG/C893 ). 35
It has recently been shown that increased levels of DQ and PA may result in substantial reductions in the risk of NAFLD and clinically significant fibrosis in a contemporary US cohort (
). Notably, benefits from increased DQ and PA may extend to both sexes and all racial/ethnic groups. However, the protection of DQ and PA on mortality in NAFLD is less clear. The purposes of this study were to examine (a) the association between PA and HEI-2015 with NAFLD (FLI≥60), (b) the dose-response association of PA and HEI-2015 with all-cause, CVD-related, and cancer-related mortality among NAFLD (FLI ≥60) participants, and (c) whether the benefits of PA and HEI-2015 on mortality might differ among participants with or without suspected NAFLD and across different ages, sex, and racial/ethnic groups. 8 METHODS
Study population and design
NHANES is a cross-sectional survey, designed to provide a representative sample of the US noninstitutionalized civilian population selected with a complex and multistage probability design (
). For these analyses, 2 consecutive cycles (2011–2012 and 2013–2014) gathering information on accelerometer-based physical activity monitoring (PAM) and food intake were selected. In these 2 cycles, older adults, Hispanic, non-Hispanic (NH) Black and Asian, and low-income NH-White/other ethnicities were intentionally oversampled. All NHANES protocols were approved by the Centers for Disease Control's National Center for Health Statistics Ethics Review Board, and all survey participants provided written informed consent ( 36 ). 36
A total of 14,693 participants aged 3 (NHANES 2012–2014) or 6 (NHANES 2011–2013) and older were asked to wear a PAM starting on the day of their examination in the mobile examination center (MEC). We initially excluded 1,193 participants who wore the device on the dominant wrist or whose wrist placement was unknown and who did not have at least 3 valid days. A valid day was defined as having 1,440 minutes of accelerometer data (i.e., 24 hours) in which less than 5% of the time was considered nonwear (i.e., <72 minutes) and less than 17 h were recorded as sleepwear (
). Furthermore, 4,263 individuals aged < 18 years and those with missing values for dietary information or who reported an implausible total energy intake (<600 or >3,500 kcal/d for female participant and <800 or >4,200 kcal/d for male participant) ( 37 ) (n = 1,294) were removed. In addition, 535 participants with excessive alcohol intake (>2 or >3 standard drinks daily in women and men, respectively) (n = 321) or positivity to hepatitis B surface antigen (n = 75), hepatitis C antibody (n = 111), and HIV (n = 28) were excluded. The final analytical data set comprised 3,495 and 3,913 individuals with or without suspected NAFLD. 38 NAFLD definition
Fatty Liver Index (FLI), an algorithm based on body mass index (BMI), waist circumference, triglycerides, and GGT, was used to identify participants with NAFLD. The FLI overall accuracy in discriminating NAFLD is 0.84 (95% CI: 0.81–0.87), and an FLI ≥60 has shown specificity of 86%–93% in different studies, including multiracial/ethnic populations (
). FLI has also been associated with all overall mortality, including cardiovascular-related and liver-related ( 39–41 ). 42,43 Physical activity assessment
PA data were collected using a wrist-worn triaxial accelerometer (
) and processed using a novel algorithm developed by machine learning with participant-level data summarized in MIMS units by minute, hour, and day ( 44 ). MIMS/min data are freely available from NHANES. In brief, participants were asked to wear an ActiGraph GT3X+ accelerometer on their nondominant wrist for 7 full and 2 partial days. Starting on the day of their MEC, participants were asked to wear the PAM for 7 consecutive days (second to eighth days of wear) and remove the device on the morning of day 9. The raw acceleration data were converted into MIMS units using a 4-step process that comprised interpolation, extrapolation, bandpass filtering, and aggregation of data for each axis, and MIMS units correspond to the total amount of movement activity ( 26 ). Daily wear time was determined through wake-wear, sleepwear, and nonwear daily estimates simultaneously calculated by a machine learning algorithm. Partial days of wear (on days 1 and 9) were removed before analyses. Previous studies indicate that for population-level analyses, between 1 and 3 days are appropriate to generate stable group-level estimates of activity ( 26 ); therefore, participants with 3 or more valid days of monitor wear time were included in this analysis. 45,46
PA volume was defined as the average of daily MIMS/min and calculated as the sum of all MIMS/min data accumulated within a day and averaged across all valid days. PA intensity (peak cadence or peak-30 MIMS/min) was defined as the average of movements/d of the highest 30 MIMS/min (not necessarily consecutive) for all valid days. Peak-30 MIMS (intensity) was obtained by (a) first rank ordering an individual's MIMS/min values within each valid day of observation, (b) calculating the mean of the highest 30 values within each day, and (c) taking the average of the resulting MIMS/min values across all valid wear days. Cadence (steps/min) as measured by direct observation and objective assessments and intensity (indirect calorimetry) has been strongly correlated (
r > 0.90). Therefore, the use of peak cadence as a marker of PA intensity may have clinical and practical significance ( ). 31
Each MIMS metric represents a particular aspect of habitual movement behavior. The average of daily MIMS/min and peak-30 MIMS/min might reflect the volume and intensity of movements, respectively. Because there are no established thresholds for MIMS metrics associated with health outcomes, we computed quartiles of MIMS metrics to examine their associations with all-cause mortality. Details on PA assessment can be found in the Supplementary Digital Content (see Supplementary Material,
). https://links.lww.com/AJG/C893 Diet quality assessment
In each NHANES cycle, participants were asked to provide detailed information on food intake for two 24-hour periods, which was then used to estimate intakes of energy, nutrients, and other dietary components. The first dietary recall was collected during the initial in-person visit, whereas the second was collected by telephone 3–10 days later. With the 2-day collected information, the HEI-2015 score was calculated (
). It is a scoring metric that can be used to determine overall DQ and the quality of several dietary components. HEI-2015 is derived from the sum of 13 components and ranges from 0 to 100. Nine components focus on adequacy (i.e., foods to eat enough of to get the nutrients needed for good health: total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and fatty acid ratio). Four components focus on moderation (i.e., foods to limit or consume in small amounts: refined grains, sodium, saturated fats, and added sugars). High HEI-2015 scores indicate better DQ. The 13 components of HEI-2015 were calculated using the total nutrient intakes on the first day, the second day, and the US Department of Agriculture MyPyramid Equivalents Database/Food Patterns Equivalents Database files. More details on the HEI-2015 score calculation can be found in the Supplementary Digital Content (see Supplementary Material, 35 ). https://links.lww.com/AJG/C893 Other participants' characteristics
Demographic characteristics (age, self-reported sex, and race/ethnicity) were queried during the home interview. Race/ethnicity was categorized as NH-White, NH-Black, NH-Asian, and other (including multiple races). Mexican Americans and other Hispanic participants were combined to create the Hispanic group. Diabetes mellitus was defined by a self-reported previous diagnosis, a glycosylated hemoglobin A1c level of ≥6.5%, or a fasting plasma glucose level of ≥126 mg/dL. Hypertension was defined based on previous physician diagnosis, taking prescribed medicine to decrease blood pressure, or blood pressure of ≥140/90 mm Hg. The BMI was calculated as the weight in kilograms divided by the height in meters squared. Smoker was defined as adults who smoked at least 100 cigarettes in life. Smokers who do not currently smoke cigarettes were considered former smokers. The history of malignant neoplasms and CVD was obtained from the medical conditions item of the survey questionnaire. Those who answered “yes” to whether they were “ever told you had cancer or malignancy” were considered cancer survivors, whereas those who answered “no” were considered noncancer survivors. The history of CVD was defined as a self-reported previous history of coronary artery disease/heart failure and/or stroke/transient ischemic attacks. Participants were considered to have a mobility limitation if they self-reported difficulty walking 0.25 miles, without special equipment, or up 10 steps. Grip strength was measured in kilograms at the NHANES MEC using a Takei dynamometer (model: TKK 5401 [Takei Scientific Instruments, Tokyo, Japan]).
All-cause mortality, and cardiovascular-related and cancer-related mortality were the main study outcomes. Ascertainment of all-cause and cause-specific mortality was performed through linkage to the National Death Index records until December 31, 2019. The cause of death was determined according to the 10th revision of the
International Classification of Disease ( ICD-10). CVD mortality was defined as ICD-10 codes I00–I09, I11, I13, I20–I51, and I60–I69, and cancer mortality was defined as ICD-10 codes C00–C97. Liver-specific mortality is not specifically identified in this cohort but is lumped into the “all other causes” category. Data analysis
Examination sample weights, accounting for nonresponse, noncoverage, and unequal selection probabilities for certain categories of the population, were incorporated in all analyses to produce national estimates. Sample weights were reweighted to account for the use of combined NHANES cycles (
). Age-adjusted estimates of participants' baseline characteristics and mortality rates (%) were calculated by the direct method to the 2000 US Census standard population, using age groups 18–39, 40–59, and 60 and older, and examination sample weights ( 47 ). Baseline characteristics of the study population were further described by quartiles of PA/DQ using means (standard errors) and percentages for continuous and categorical variables, respectively. Standard errors were calculated with the Taylor series linearization method ( 48 ). Comparisons of baseline characteristics by PA/DQ quartiles were performed using linear regression for continuous variables and the design-adjusted Rao-Scott χ 47 2 test for categorical variables. No missing data for our main covariates and outcomes were recorded in our data set.
First, we tested univariate and covariate-adjusted associations between PA/DQ and NAFLD (FLI≥60) using logistic regression models. Second, we assessed a dose-response association between PA/DQ metrics and all-cause mortality using restricted cubic splines using Cox regression models. Knots were placed at the 5th, 35th, 65th, and 95th percentiles of the exposure distribution. We used the
testparm Stata command that uses a Wald test-type test with 2 degrees of freedom to examine linearity. A P value < 0.05 indicated strong evidence against linearity. We estimated the minimal and optimal doses of each metric at which the minimum and maximum significant risk reductions were observed. E values associated with the minimal and optimal doses were calculated to estimate the likelihood of bias from unmeasured confounding ( ). The higher the E value is, the stronger the unmeasured confounding must be to explain the observed association. Finally, population attributable fractions (PAF) were used to estimate how much mortality could be prevented if, at the time of study enrollment, all individuals would have adopted very high levels of PA/DQ, which correspond with values within the top or 4th quartile. PAF were calculated with the 49,50 punafcc package (developed for survival data) after fitting our fully covariate-adjusted Cox survival regression models ( ). More information on PAF can be found in the Supplementary Digital Content (see Supplementary Material, 51,52 ). https://links.lww.com/AJG/C893
Cox proportional hazards regression models were conducted to examine the association of PA/DQ with overall mortality by calculating hazard ratios (HR) and 95% confidence intervals (CI). Competing risk regression models were performed to determine the association of PA/DQ with cause-specific mortality by calculating subhazard ratios (sHR) and 95% CI. Cardiovascular-related (heart disease and cerebrovascular disease) and cancer-related mortality were the main causes to be analyzed. Other causes of death than CVD and cancer-related were considered competing events.
Two models were used to examine the association of PA/DQ with all-cause and cause-specific mortality. In model 1, we only adjusted for age, sex, and racial/ethnic groups. In model 2, we further adjusted for BMI, smoking status, diabetes mellitus, hypertension, history of CVD or cancer, total calorie intake (kcal), hand grip strength, and mobility limitation. PA/DQ metrics were mutually included in models 1 and 2. These covariates were selected based on their strong relationships with mortality, NAFLD, and PA/DQ levels. No violations to the proportional hazards assumption were noticed using the Schoenfeld residual plots.
Multiple sensitivity analyses were performed to explore associations between PA/DQ and mortality across the following groups: (a) participants with an FLI <60 (non-NAFLD), (b) ages <40, 45–60, and ≥65 years, (c) sex (female and male), (d) race/ethnicity (non-Hispanic White, non-Hispanic Black, and Hispanic), and (e) excluding participants who died within the first year of follow-up to deal with the possible effect of reverse causation. We also performed additional analyses to examine the joint effects of PA (volume and intensity) and DQ on the risk of all-cause mortality. To do so, we refit the fully adjusted all-cause mortality models by including 2-way interaction terms between quartile indicators for PA (volume and intensity) and HEI-2015. The lowest quartile of PA volume, PA intensity, or HEI-2015 was considered the referent group.
To estimate the adequate sample size of our study, we computed the effective power using the number of events per variable, by dividing the number of deaths by the number of variables considered in our analyses and compared it against a critical value of 10. The events per variable in this study was at least 23 for all analyses, meaning the study was adequately powered (
). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines for cross-sectional studies ( 53 ). Statistical analyses were performed using R software version 4.0.2 (R Foundation for Statistical Computing) and Stata MP version 17.0 (StataCorp, College Station, TX), and a 2-sided significance level of 0.05 was used. 54 RESULTS
Association of PA (volume and intensity) and DQ (HEI-2015) with NAFLD (FLI≥60)
Inverse and dose-dependent relationships were observed between quartiles of PA (volume and intensity) and HEI-2015 and the risk of NAFLD (
Table 1). In fully covariate-adjusted logistic regression models, the top quartile of PA volume (odds ratio [OR]: 0.50, 95% CI: 0.39–0.63), PA intensity (OR: 0.68, 95% CI: 0.57–0.82), or HEI-2015 (OR: 0.46, 95% CI: 0.39–0.53) was associated with the lowest risk of NAFLD. Table 1.:
Association of PA (volume and intensity) and HEI-2015 quartiles with NAFLD
All-cause mortality by NAFLD status
In the NAFLD population (n = 3,495), the mean follow-up duration was 6.79 years, with a maximum of 9 years. There were 298 (age-adjusted percentage = 6.4%) deaths over 23,723 person-years of follow-up (mortality rate of 1.26 deaths/100 person-years). Among participants without NAFLD (n = 3,913), the mean follow-up duration was 6.83 years, with a maximum of 9 years. There were 264 (age-adjusted percentage = 4.9%) deaths over 26,756 person-years of follow-up (mortality rate of 0.98 deaths/100 person-years). As compared with non-NAFLD, NAFLD was associated with an increased risk of all-cause mortality (adj. HR: 1.26, 95% CI: 1.05–1.52) (see Supplementary Table 1,
). https://links.lww.com/AJG/C893 Baseline characteristics by PA and DQ in the NAFLD cohort
Age-adjusted baseline characteristics by quartiles of PA (volume and intensity) and DQ are summarized in the Supplementary Digital Content (see Supplementary Tables 2 and 3,
). Higher quartiles of PA (volume and intensity) were associated with younger age, female sex, Hispanic, lower BMIs/waist circumferences, lower prevalence of diabetes, hypertension, cardiovascular conditions, and cancer, and higher DQ. A higher DQ was observed in older adults, women, Hispanic, and NH-Asian. Participants in the upper or third quartile of HEI-2015 had lower BMIs/waist circumferences and were less likely to smoke and have hypertension, diabetes, CVD, and malignancies. https://links.lww.com/AJG/C893 Associations between PA (volume and intensity), DQ, and mortality in the NAFLD cohort
Age-adjusted baseline characteristics by mortality status in the NAFLD cohort are summarized in
Table 2. Overall, the upper or third quartile of PA (volume and intensity) and DQ was associated with lower rates of all-cause mortality. Table 2.:
Baseline characteristics by all-cause mortality status in the NAFLD population
Effect of PA (volume and intensity) on mortality
In fully covariate-adjusted Cox models, higher levels of PA, including total volume and intensity, were associated with a lower risk of all-cause mortality. The covariate-adjusted cubic restricted spline analysis showed a negative and nonlinear (Wald test
P value < 0.05) dose-dependent association between PA volume and all-cause mortality, wherein the maximum protective dose was observed at 14,300 MIMS/min (adj. HR: 0.20, 95% CI: 0.11–0.38) ( Figure 1a), E value: 5.19 (CI: 2.7). The intensity of PA was also negatively independently associated with all-cause mortality in a nonlinear (Wald test P value = 0.05) and dose-dependent manner. The maximum protective dose was observed at 54.25 MIMS/min (adj. HR: 0.10, 95% CI: 0.05–0.23) ( Figure 1c), E value: 12.79 (CI: 5.9). We did not observe minimal thresholds for beneficial effects of PA volume/intensity on all-cause mortality. Figure 1b,d shows the relative frequency of PA volume and intensity in the NAFLD population. Of note, 30% and 5% of participants achieved the maximum thresholds of PA volume and intensity associated with survival benefits. Cubic restricted spline analyses in the population that excluded those participants who died (n = 18) during the first year of follow-up showed quite similar results (see Supplementary Figures 1A and 1B, ). https://links.lww.com/AJG/C891 Figure 1.:
Cox regression-based cubic splines representation of the relationship between PA (volume and intensity), HEI-2015, and all-cause mortality in the NAFLD population. (
a) PA volume (MIMS/min), ( b) distribution of PA volume (MIMS/min), ( c) PA intensity (MIMS/min), ( d) distribution of PA intensity (MIMS/min), ( e) HEI-2015, and ( f) distribution of HEI-2015. PA volume: The maximum benefit on mortality was seen at 14,300 MIMS/min (HR: 0.20, 95% CI: 0.11–0.38). PA intensity: The maximum benefit on mortality was seen at 54.25 MIMS/min (HR: 0.10, 95% CI: 0.05–0.23). HEI-2015: The benefit on mortality was noticeable at an HEI-2015 score of 48.12 (HR: 0.87, 95% CI: 0.76–0.99), whereas the maximum benefit on mortality was seen at an HEI-2015 score of 66.17 (HR: 0.54, 95% CI: 0.40–0.74). HEI, Healthy Eating Index; MIMS, Monitor-Independent Movement Summary; NAFLD, nonalcoholic fatty liver disease; PA, physical activity.
We further analyzed the association of PA (volume and intensity), stratified by quartiles, with all-cause and CVD-related and cancer-related deaths (
Table 3). Compared with participants in the lowest quartile of the total volume of PA (9,882 MIMS/min), those who performed >14,867 MIMS/min (top quartile) displayed the lowest risk for all-cause (adj. HR: 0.45, 95%: 0.30–0.68) and CVD-related (adj. sHR: 0.43, 95% CI: 0.19–0.98) deaths. Participants in the top quartile of peak-30 MIMS/min (>45.13) showed the lowest risk of all-cause (adj. HR: 0.39, 95% CI: 0.23–0.66) and CVD-related (adj. sHR: 0.26, 95% CI: 0.11–0.63) deaths compared with those in the first quartile (<36.56). Interestingly, neither PA volume nor intensity was independently associated with cancer-related mortality. Table 3.:
Risk of all-cause, and cardiovascular-related and cancer-related mortality by PA and HEI-2015 in the NAFLD population
The protective effects of PA volume and intensity on all-cause and CVD mortality risk remained statistically significant in the sensitivity analysis removing participants who died within the first 1 year of follow-up (see Supplementary Table 4,
). Minimally adjusted (age, race/ethnicity, sex, and DQ) Cox models examining the association between PA and all-cause mortality by sex, race/ethnicity, and age (<45, 45–65, and >65 years) showed that those in the top quartile of PA volume (see Supplementary Table 5, https://links.lww.com/AJG/C893 ) and intensity (see Supplementary Table 6, https://links.lww.com/AJG/C893 ) had a reduced mortality risk compared with those in the lowest tertile (adj. Cox model trend https://links.lww.com/AJG/C893 P value < 0.05 in all subgroup analyses). Diet quality (HEI-2015)
We found a negative, linear (Wald test
P value > 0.05), and dose-dependent association between DQ and all-cause mortality in the fully covariate-adjusted cubic restricted spline analysis ( Figure 1e), wherein the minimal and maximum protective effects were seen at HEI-2015 score cutoffs of 48.12 (adj. HR: 0.87, 95% CI: 0.76–0.99) and 66.17 (adj. HR: 0.54, 95% CI: 0.40–0.74), E values: 1.79 (CI: 1.01) and 2.99 (CI: 2.01), respectively. The relative frequency of the HEI-2015 score across the NAFLD population is represented in Figure 1f. The proportion of participants with HEI-2015 scores of ≥48.12 and ≥66.17 was 63% and 17%. After removing participants who died within the first year of follow-up, the cubic restricted spline analysis displayed the same pattern of association between DQ and all-cause mortality (see Supplementary Figure 1C, ). The analysis using DQ stratified by quartiles showed that participants with an HEI cutoff of ≥61.87 (top quartile) had the lowest risk for all-cause (adj. HR: 0.64, 95% CI: 0.45–0.89), CVD-related (adj. sHR: 0.69, 95% CI: 0.38–0.99), and cancer-related (adj. sHR: 0.45, 95% CI: 0.22–0.92) mortality as compared to those in the lowest quartile ( https://links.lww.com/AJG/C891 Table 3). The protection conferred by DQ on the risk of mortality remained statistically significant even after removing participants who died within the first 1 year of follow-up (see Supplementary Table 4, ). The Supplementary Digital Content (see Supplementary Table 7, https://links.lww.com/AJG/C893 ) illustrates associations between DQ and all-cause mortality across ages, sexes, and racial/ethnic groups. The risk of all-cause mortality remained significantly statistically lower in participants displaying higher levels of DQ as compared to those within the first quartile across all population subgroups (P for trend across all subgroups <0.05). https://links.lww.com/AJG/C893 Table 4 shows associations of adequacy and moderation of HEI-2015 food components with all-cause mortality. Survivors showed significantly statistically higher consumption of total vegetables, greens and beans, total and whole fruits, total protein foods, seafood and plant proteins, and lower consumption of added sugar; P value < 0.05 after controlling for age, sex, race/ethnicity, and total caloric intake. Table 4.:
Association of Healthy Eating Index (2015) component scores with mortality in the NAFLD population
Joint effects of PA (volume and intensity) and DQ on the risk of all-cause mortality
There were significant interactions between PA volume and intensity, PA volume and DQ, and PA intensity and DQ and risk of all-cause mortality, all
P values for those interactions <0.01. Higher levels of peak-30 MIMS/min (intensity) remained significantly associated with lower mortality risk even in participants displaying very low levels of PA volume (<9,882 MIMS/min). Participants engaging in the highest levels of PA volume (>14,867 MIMS/min) and intensity (>45.13 MIMS/minute) showed the highest protection against all-cause mortality (HR: 0.23, 95% CI: 0.14–0.53), Figure 2a. Participants adhering to the healthiest eating patterns showed lower mortality rates even among those with low levels of PA volume (<9,882 MIMS/min). Of note, the maximum survival benefit was seen in those reporting the highest levels of DQ (>61.87) and PA volume (>14,867), HR: 0.09, 95% CI: 0.02–0.28, Figure 2b. Finally, PA intensity was also significantly linked to lower mortality rates regardless of DQ levels. However, survival benefits linked to higher levels of PA intensity were more evident in those participants adopting healthier eating patterns. For example, the HR for all-cause mortality among people with a PA intensity of >45.13 and an HEI of >61.87 was 0.05 (95% CI: 0.01–0.36), Figure 2c. Figure 2.:
Joint effects of PA (volume and intensity in MIMS/min) and HEI-2015 on the risk of all-cause mortality in the NAFLD population. (
a) Physical activity intensity-by-volume. ( b) HEI-2015-by-physical activity volume. ( c) Physical activity intensity-by-HEI-2015. Interaction terms between PA intensity-by-PA volume, HEI-2015-by-PA volume, and PA intensity-by-HEI-2015 were significant ( P < 0.01). For all analyses, the lowest quartile of PA volume, PA intensity, or HEI-2015 was considered the referent group. HEI, Healthy Eating Index; HR, hazard ratio; MIMS, Monitor-Independent Movement Summary; NAFLD, nonalcoholic fatty liver disease; PA, physical activity. Beneficial effects of PA (volume and intensity) and DQ on the risk of all-cause mortality in participants without NAFLD
The fully covariate-adjusted cubic restricted spline analysis showed the same patterns of association between PA (volume and intensity) and HEI-2015 and all-cause mortality in non-NAFLD as compared with NAFLD.
Figure 3a,c,e displays dose-response associations between PA/DQ metrics and all-cause mortality, and Figure 3b,d,f shows relative frequencies of PA/DQ metrics in the non-NAFLD population. Overall, there was no evidence of linearity when applying restricted cubic splines to the association between PA volume ( P < 0.01) or intensity ( P < 0.01) and risk of all-cause mortality. However, we found a negative and linear (Wald test P > 0.05) dose-dependent association between DQ and risk of all-cause mortality. Of note, the maximum protective dose for PA volume or intensity was observed at 23,000 MIMS/min (adj. HR: 0.03, 95% CI: 0.01–0.18) and 70.35 MIMS/min (adj. HR: 0.02, 95% CI: 0.01–0.08). Like the NAFLD population, we did not observe minimal thresholds for the beneficial effects of PA volume/intensity on mortality risk. Regarding DQ, the benefit on mortality risk started at an HEI-2015 score of 65.64 (HR: 0.70, 95% CI: 0.50–0.98), whereas the maximum benefit on mortality risk was seen at an HEI-2015 score of 96.02 (HR: 0.17, 95% CI: 0.07–0.41). Figure 3.:
Cox regression-based cubic spline representation of the relationship between PA (volume and intensity), HEI-2015, and all-cause mortality in the non-NAFLD population. (
a) PA volume (MIMS/min). ( b) Distribution of PA volume (MIMS/min). ( c) PA intensity (MIMS/min). ( d) Distribution of PA intensity (MIMS/min). ( e) HEI-2015. ( f) Distribution of HEI-2015. PA volume: The maximum benefit on mortality was seen at 23,000 MIMS/min (HR: 0.03, 95% CI: 0.01–0.18). PA intensity: The maximum benefit on mortality was seen at 70.35 MIMS/min (HR: 0.02, 95% CI: 0.01–0.08). HEI-2015: The benefit on mortality starts at an HEI-2015 score of 65.64 (HR: 0.70, 95% CI: 0.50–0.98), whereas the maximum benefit on mortality is seen at an HEI-2015 score of 96.02 (HR: 0.17, 95% CI: 0.07–0.41). HEI, Healthy Eating Index; MIMS, Monitor-Independent Movement Summary; NAFLD, nonalcoholic fatty liver disease; PA, physical activity. Table 5 shows how different are the beneficial effects of PA/DQ metrics on the risk of all-cause mortality among participants with or without NAFLD. The beneficial effects of PA/DQ metrics were compared using HR and PAF. Overall, the protection conferred by PA (volume and intensity) on mortality risk seemed higher in those without NAFLD vs with NAFLD. Among non-NAFLD participants, those reporting either PA volume of >15,407 MIMS/min or intensity of >47.21 MIMS/min displayed higher risk reductions in mortality (47% or 65%) as compared with those carrying a diagnosis of NAFLD (40% or 50%). The PAF indicate that increasing PA volume beyond 15,407 MIMS/min would be associated with a 43% and 36% decrease in the number of deaths in the non-NAFLD and NAFLD cohorts. The PAF also suggest that an increase in PA intensity beyond 47.21 MIMS/min would be associated with the prevention of 62% and 48% of total deaths among non-NAFLD and NAFLD individuals. Finally, we estimated that benefits attributed to adopting the healthiest dietary patterns (HEI >63.53) on mortality risk were equal among individuals without or with NAFLD. Table 5.:
Beneficial effects of physical activity (volume and intensity) and diet quality (HEI-2015) on the risk of all-cause mortality
Supplementary Digital Content (see Supplementary Table 8,
) displays baseline characteristics for participants with or without NAFLD. As compared with NAFLD, the non-NAFLD group consisted of younger, less obese, fewer diabetic, and hypertensive individuals. https://links.lww.com/AJG/C893
Supplementary Digital Content (see Supplementary Table 9,
) shows a further analysis examining a multiplicative interaction term between NAFLD-by-PA (volume in MIMS/min) or NAFLD-by-PA (intensity in MIMS/min) or NAFLD-by-HEI-2015 and risk of all-cause mortality. In addition to our results shown in https://links.lww.com/AJG/C893 Table 5, this analysis confirms that the protection conferred by PA (both modalities) on the risk of all-cause mortality was statistically significantly higher in non-NAFLD compared with NAFLD participants. DISCUSSION
In a cohort of US adults, we showed an inverse association of PA (volume and intensity) as determined by MIMS/min and DQ by HEI-2015 with the risk of NAFLD. In the NAFLD population, either the volume or intensity of PA was negatively associated with the risk of all-cause and CVD-related mortality in a dose-dependent manner. Furthermore, higher adherence to a healthy eating pattern as assessed by HEI-2015 was linked to a lower risk of all-cause, CVD-related, and cancer-related mortality. Of the 13 HEI-2015 components, scores of total vegetables, greens and beans, total and whole fruits, total protein foods, seafood and plant proteins, and added sugar were the most associated with mortality. The observed associations seemed not to be modified by other risk factors with a recognized link to mortality, PA, or DQ. More importantly, benefits from increased PA and DQ extended to young, middle age, and older participants as well as both sexes and all racial and ethnic groups, excluding NH-Asian (not analyzed because of a low number of reported deaths). Interestingly, either PA (volume and intensity) or DQ showed the same patterns of protection against mortality in people without NAFLD, although PA-related survival benefits seemed to be more clinically relevant in the non-NAFLD than the NAFLD population. Taken together, these findings suggest the potential for lifestyle modification through PA and DQ to reduce mortality risk and disease burden associated with NAFLD.
Physical activity and mortality
To the best of our knowledge, this is the first study examining the effect of total volume and intensity of PA as assessed by a worn-wrist accelerometer, with participant-level data summarized in MIMS units by minute, on the risk of mortality in a contemporary US population with suspected NAFLD.
Our data showed nonlinear associations of PA volume and intensity with all-cause mortality. We did not observe minimal thresholds for the beneficial effects of PA volume/intensity on all-cause mortality, supporting the notion that even small amounts of PA may provide important benefits on health, including all-cause mortality. This could suggest the need for a contemporary paradigm shift in public health recommendations for PA, which recommends MVPA as an important lifestyle behavior regardless of the bout duration (
). We estimated the maximum dose of PA volume at approximately 14,300 MIMS/min per day, which was associated with an 80% risk reduction in all-cause mortality. PA intensity (peak 30-minute cadence), which may reflect a person's best natural effort in a free-living environment ( 55 ), had a remarkable protective effect on all-cause mortality. When analyzing the spline graphic, we observed a steady decline in mortality risk with more peak-30 MIMS/min (intensity) up to approximately 54.25 MIMS/min, beyond which mortality risks flattened. At a threshold of ≥54.25 peak-30 MIMS/min, the mortality risk reduction was remarkably high (90%). Our analysis stratifying PA (volume and intensity) by quartiles showed that both volume and intensity of PA may confer significant protection against all-cause and CVD-related deaths, the higher the volume and intensity and the greater the benefits on mortality risk. To better understand the joint effects of PA volume and intensity on the risk of mortality, we examined the benefits of peak-30 MIMS/min across different levels of total PA volume. This analysis revealed that peak-30 MIMS/min, as a proxy of intensity, was independently associated with a lower risk of mortality across all volumes of PA. Interestingly, the benefit was even extended to participants with lower PA volume. This finding indicates that both daily total volume and intensity (peak effort) are strongly and independently associated with better survival rates in NAFLD. 31
Despite PA levels having been consistently associated with better survival rates among cancer survivors, we found no protective effect of PA on cancer-related mortality (
). However, our findings should be interpreted with caution for different reasons: (a) details on mortality related to specific types of cancers were not available in the NHANES data, and PA may have beneficial effects on survival for patients with breast, colorectal, and prostate cancers ( 56,57 ), but the evidence is very limited for other cancers, and (b) information of clinical stages of cancers and types of treatments participants were receiving at study entry was unavailable. Clinically advanced cancers and acute side effects of cancer treatment modalities may influence both ability and desire to participate in PA and therefore confound the association between PA and cancer-specific mortality. 57
Our data are alignment with a previous population-based study including participants with NAFLD from the NHANES cycle 2003–2006 and examining the effect of the total duration of PA and MVPA as determined by a hip-worn accelerometer on the risk of mortality (
). This study showed that total PA and MVPA may confer protection against all-cause mortality. The total duration of PA, but not MVPA, showed clear protection against cardiovascular mortality. Neither the total duration of PA nor MVPA was associated with a lower risk of death due to cancers. Despite the potential protective effects of higher levels of MVPA on mortality, our findings might not be comparable to the previous study using accelerometer-measured MVPA. First, each MIMS metric represents a particular aspect of habitual movement behavior. Overall daily MIMS may represent PA volume irrespective of intensity, which may differ from self-reported MVPA or accumulation of activity minutes above the previously established MVPA cutoffs. Second, in contrast to MVPA cut points that underestimate light PA intensity, peak-30 MIMS/minute reflects a wide spectrum of acceleration magnitudes ranging from lower to higher peak efforts within a day and across all valid days, facilitating the comparison across a full range of PA intensity levels. This is notably important for older adults because this population tends to spend relatively little time above MVPA cutoffs ( 58 ). 28,29
Finally, our data show that both modalities of PA remained significantly associated with a lower risk of mortality regardless of the NAFLD status. However, both PA volume and intensity led to greater survival benefits in people without NAFLD compared with NAFLD (see Supplementary Table 9,
). It is not clear why participants with suspected NAFLD derive less benefit from PA, but it could be related to older age, higher BMI, and greater prevalence of metabolic factors such as diabetes and hypertension (see Supplementary Table 8, https://links.lww.com/AJG/C893 ). https://links.lww.com/AJG/C893 Diet quality and mortality
We observed a linear and inverse dose-dependent association between DQ and all-cause mortality. The spline analysis suggested that a minimum threshold of 48.12 was required to lower the risk of mortality. The maximum dose was observed at approximately 66.17, which was associated with a 46% risk reduction in mortality. Our analysis stratifying DQ by quartiles showed that the highest quartile did confer significant protection against all-cause and CVD-specific and cancer-specific mortality, even after controlling for relevant demographic, comorbidities, and lifestyle factors including PA volume and intensity. The inverse association with mortality was apparent in both sexes and at different strata of age and races/ethnicities. Our results are in alignment with several studies reporting the positive impact of various indexes reflecting DQ on all-cause and cause-specific mortality including those due to CVD and cancers in the general and NAFLD population (
Further analysis of the association between individual dietary components and mortality showed that mean consumption levels of total vegetables, greens and beans, total and whole fruits, total protein foods, and seafood and plant protein were significantly higher in those who remained alive vs those deceased. By contrast, a higher level of added sugar intake was observed among deceased vs alive participants. Concordant with the present findings, previous cohort studies found an inverse association between higher vegetable and plant-based protein intake and all-cause, CVD, and cancer mortality (
). Other studies have also reported the deleterious effects of added sugar on all-cause, CVD-specific, and cancer-specific deaths ( 63,64 ). Collectively, these data suggest that a higher intake of vegetables, greens and beans, seafood and plant proteins, and fruits but a lower intake of sugar may exert beneficial effects on all-cause mortality, and these results reinforce the concept that the quality of what we eat may have profound effects on health outcomes and support the current DGA recommendations about the importance of healthy eating patterns as a whole instead of focusing on individuals' nutrients or foods. 65–67
DQ was also significantly associated with lower all-cause mortality in participants without NAFLD. Of note, patterns of association between DQ and mortality were quite similar to those observed in participants with NAFLD.
This study had several strengths. We used nationally representative data from a large sample of US participants with possible NAFLD. We examined the independent associations between PA and DQ with all-cause and cause-specific mortality. Moreover, our study explored for the first time the association between the novel PA metric (MIMS/min) and mortality. The data we report herein may serve as benchmark values related to PA (MIMS volume and intensity metrics), DQ (HEI-2015), and mortality in NAFLD adults, although our data need to be reproduced or validated in other cohorts including adults from general and NAFLD populations. This study also extended the results of previous research, adding more evidence of the potential benefits of PA and DQ in a clinical context.
There are important limitations that we also recognize. MIMS/min is a new measure of PA with no cutoffs available to classify total volume and intensity that meet PA guidelines. It is yet unknown how well MIMS/min will predict energy expenditure, and to date, there are no verified thresholds to allow the MIMS to be converted to time spent in different activity intensity categories that are strongly associated with health outcomes. Our analysis provides a first step toward the use of a standardized metric to link activity behaviors with mortality in a nationally representative population. Therefore, the practical relevance of the current PA cut-off points given in MIMS/min may be validated in external cohorts. NAFLD was classified according to a noninvasive maker (FLI), and it could be associated with potential misclassification. Measurement errors in self-reported diet, lifestyle factors, and comorbid conditions (i.e., diagnosis of CVD or cancer at baseline) were inevitable. Reverse causation and residual confounding may still be present. However, the large E values for the association between both modalities of PA and DQ with mortality might suggest that unmeasured confounders were unlikely to explain the association between PA/DQ and mortality. Because liver-related deaths are not reported in publicly available National Death Index files, we could not examine the association between DQ/PA and liver-specific mortality. Finally, there is the potential for unmeasured confounding, particularly those linked to the severity of comorbid conditions, functional status, acute events leading to hospitalizations, and type of medications received, which could be associated with the mortality outcomes in this study.
In conclusion, among US adults with NAFLD, increased adherence to healthier eating patterns and increased total volume and intensity of PA were associated with lower risk of all-cause and CVD-specific mortality. Adhering to healthier eating patterns also conferred protection against cancer-related mortality. Survival benefits achieved from increased levels of DQ and PA were also seen in participants without NAFLD. Notably, the current study will help guide our understanding of PA MIMS/min metric thresholds aligned with survival rates using free-living wrist-worn accelerometry data from a nationally representative sample (NHANES).
CONFLICTS OF INTEREST
Guarantor of the article: Naga Chalasani, MD, and Eduardo Vilar-Gomez, MD, PhD.
Specific author contributions: All authors made substantial contributions to the intellectual content of the paper and approved the final version of the manuscript. Conception and design: N.C. and E.V.-G. Acquisition of data: E.V.-G. Analysis and interpretation of data: N.C., E.V.-G., R.V., S.G., F.P., and N.S. Drafting of the manuscript: E.V.-G., N.C., R.V., S.G., F.P., and N.S. Critical revision of the manuscript for intellectual content: N.C., E.V.-G., R.V., S.G., F.P., and N.S. Statistical analysis: E.V.-G. and F.P. Obtaining funding: N.C. Supervision: N.C. and E.V.-G.
Financial Support: This study was funded by departmental internal funding and did not influence the study's design, conduct, or reporting. Potential competing interests: There are none for this paper. N.C. has ongoing paid consulting activities (or had in the preceding 12 months) with AbbVie, Madrigal, Foresite labs, Altimmune, Zydus, Galectin, GSK, Lilly, and Boehringer-Ingelheim. N.C. receives research grant support from Exact Sciences and DSM, where his institution receives the funding. N.C. has equity ownership in RestUp, Inc. S.G. discloses consulting for TransMedics and Pfizer and receives research grant support from Zydus, Galmed, Viking, and Sonic Incytes. R.V. receives research grant support from Zydus, Galectin Therapeutics, Gilead Sciences, and Novo Nordisk, Eli Lilly, and also discloses consulting for Labcorp, Medpace, GSK, Daiichi Sankyo, and Echosens. E.V.-G., F.P., and N.S. have nothing to disclose.
Study Highlights WHAT IS KNOWN
✓ Physical activity (PA) and diet quality (DQ) are modifiable risk factors shown to be strongly associated with nonalcoholic fatty liver disease (NAFLD).
✓ Cardiovascular diseases and malignant neoplasms have been identified as the commonest causes of death in NAFLD.
✓ The benefits of PA/DQ on all-cause, cardiovascular-related, and cancer-related mortality have been poorly documented in NAFLD.
WHAT IS NEW HERE
✓ DQ was assessed by the most recent iteration of the Healthy Eating Index (2015).
✓ PA was measured by wrist-worn accelerometers and expressed in a novel metric (Monitor-Independent Movement Summary units).
✓ DQ was inversely associated with all-cause, cardiovascular-related, and cancer-related mortality in a linear and dose-dependent manner.
✓ Higher levels of PA volume and intensity were associated with a lower risk of all-cause and cardiovascular mortality in a nonlinear and dose-dependent manner.
✓ The PA and DQ survival benefits were extended to participants without NAFLD and across different ages, sex, and racial/ethnic groups.
The authors thank the members of the National Center for Health Statistics, and Centers for Disease Control and Prevention, for gathering the data and making it available for public use. The authors also thank the participants involved in the survey.
1. Chalasani N, Younossi Z, Lavine JE, et al. The diagnosis and management of nonalcoholic fatty liver disease: Practice guidance from the American Association for the Study of Liver Diseases. Hepatology 2018;67(1):328–57.
2. Younossi Z, Anstee QM, Marietti M, et al. Global burden of NAFLD and NASH: Trends, predictions, risk factors and prevention. Nat Rev Gastroenterol Hepatol 2018;15(1):11–20.
3. Yip TCF, Vilar-Gomez E, Petta S, et al. Geographical similarity and differences in the burden and genetic predisposition of NAFLD. Hepatology 2022.
4. Cotter TG, Rinella M. Nonalcoholic fatty liver disease 2020: The state of the disease. Gastroenterology 2020;158(7):1851–64.
5. Sanyal AJ, Van Natta ML, Clark J, et al. Prospective study of outcomes in adults with nonalcoholic fatty liver disease. N Engl J Med 2021;385(17):1559–69.
6. Hagström H, Nasr P, Ekstedt M, et al. Fibrosis stage but not NASH predicts mortality and time to development of severe liver disease in biopsy-proven NAFLD. J Hepatol 2017;67(6):1265–73.
7. Vilar-Gomez E, Vuppalanchi R, Gawrieh S, et al. CAP and LSM as determined by VCTE are independent predictors of all-cause mortality in the US adult population. Hepatology 2023. doi: 10.1097/HEP.0000000000000023. [Epub ahead of print Janauary 3, 2023.]
8. Vilar-Gomez E, Nephew LD, Vuppalanchi R, et al. High-quality diet, physical activity, and college education are associated with low risk of NAFLD among the US population. Hepatology 2022;75(6):1491–506.
9. Vilar-Gomez E, Martinez-Perez Y, Calzadilla-Bertot L, et al. Weight loss through lifestyle modification significantly reduces features of nonalcoholic steatohepatitis. Gastroenterology 2015;149(2):367–78.e5.
10. Gerber L, Otgonsuren M, Mishra A, et al. Non-alcoholic fatty liver disease (NAFLD) is associated with low level of physical activity: A population-based study. Aliment Pharmacol Ther 2012;36(8):772–81.
11. Kim D, Vazquez-Montesino LM, Li AA, et al. Inadequate physical activity and sedentary behavior are independent predictors of nonalcoholic fatty liver disease. Hepatology 2020;72(5):1556–68.
12. Orci LA, Gariani K, Oldani G, et al. Exercise-based interventions for nonalcoholic fatty liver disease: A meta-analysis and meta-regression. Clin Gastroenterol Hepatol 2016;14(10):1398–411.
13. Hashida R, Kawaguchi T, Bekki M, et al. Aerobic vs. resistance exercise in non-alcoholic fatty liver disease: A systematic review. J Hepatol 2017;66(1):142–52.
14. Stine JG, Schreibman IR, Faust AJ, et al. NASHFit: A randomized controlled trial of an exercise training program to reduce clotting risk in patients with NASH. Hepatology 2022;76(1):172–85.
15. van Kleef LA, Hofman A, Voortman T, et al. Objectively measured physical activity is inversely associated with nonalcoholic fatty liver disease: The rotterdam study. Am J Gastroenterol 2022;117(2):311–8.
16. Zhang X, Goh GB, Chan WK, et al. Unhealthy lifestyle habits and physical inactivity among Asian patients with non-alcoholic fatty liver disease. Liver Int 2020;40(11):2719–31.
17. Samitz G, Egger M, Zwahlen M. Domains of physical activity and all-cause mortality: Systematic review and dose-response meta-analysis of cohort studies. Int J Epidemiol 2011;40(5):1382–400.
18. Moore SC, Lee IM, Weiderpass E, et al. Association of leisure-time physical activity with risk of 26 types of cancer in 1.44 million adults. JAMA Intern Med 2016;176(6):816–25.
19. Morze J, Danielewicz A, Hoffmann G, et al. Diet quality as assessed by the healthy eating index, alternate healthy eating index, dietary Approaches to Stop hypertension score, and health outcomes: A second update of a systematic review and meta-analysis of cohort studies. J Acad Nutr Diet 2020;120(12):1998–2031.e15.
20. Bonaccio M, Di Castelnuovo A, Costanzo S, et al. Mediterranean diet and mortality in the elderly: A prospective cohort study and a meta-analysis. Br J Nutr 2018;120(8):841–54.
21. Reedy J, Krebs-Smith SM, Miller PE, et al. Higher diet quality is associated with decreased risk of all-cause, cardiovascular disease, and cancer mortality among older adults. J Nutr 2014;144(6):881–9.
22. Sotos-Prieto M, Bhupathiraju SN, Mattei J, et al. Association of changes in diet quality with total and cause-specific mortality. New Engl J Med 2017;377(2):143–53.
23. Schuna JM Jr, Johnson WD, Tudor-Locke C. Adult self-reported and objectively monitored physical activity and sedentary behavior: NHANES 2005-2006. Int J Behav Nutr Phys Act 2013;10(1):126.
24. Schrack JA, Zipunnikov V, Goldsmith J, et al. Assessing the “physical cliff”: Detailed quantification of age-related differences in daily patterns of physical activity. J Gerontol A Biol Sci Med Sci 2014;69(8):973–9.
25. Tudor-Locke C, Barreira TV, Schuna JM Jr, et al. Improving wear time compliance with a 24-hour waist-worn accelerometer protocol in the international study of childhood obesity, lifestyle and the environment (ISCOLE). Int J Behav Nutr Phys Act 2015;12(1):11.
26. John D, Tang Q, Albinali F, et al. An open-source monitor-independent movement summary for accelerometer data processing. J Meas Phys Behav 2019;2(4):268–81.
27. Belcher BR, Wolff-Hughes DL, Dooley EE, et al. US population-referenced percentiles for wrist-worn accelerometer-derived activity. Med Sci Sports Exerc 2021;53(11):2455–64.
28. TROIANO RP, BERRIGAN D, DODD KW, et al. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc 2008;40(1):181–8.
29. Tudor-Locke C, Brashear MM, Katzmarzyk PT, et al. Peak stepping cadence in free-living adults: 2005–2006 NHANES. J Phys Activity Health 2012;9(8):1125–9.
30. Tudor-Locke C, Aguiar EJ. Toward comprehensive step-based physical activity guidelines: Are we ready? Kinesiol Rev 2019;8(1):25–31.
31. Tudor-Locke C, Han H, Aguiar EJ, et al. How fast is fast enough? Walking cadence (steps/min) as a practical estimate of intensity in adults: A narrative review. Br J Sports Med 2018;52(12):776–88.
32. Paluch AE, Gabriel KP, Fulton JE, et al. Steps per day and all-cause mortality in middle-aged adults in the coronary artery risk development in young adults study. JAMA Netw Open 2021;4(9):e2124516.
33. TUDOR-LOCKE C, SCHUNA JMJ, HAN H, et al. Step-based physical activity metrics and cardiometabolic risk: NHANES 2005–2006. Med Sci Sports Exerc 2017;49(2):283–91.
34. Krebs-Smith SM, Pannucci TE, Subar AF, et al. Update of the healthy eating index: HEI-2015. J Acad Nutr Diet 2018;118(9):1591–602.
35. Reedy J, Lerman JL, Krebs-Smith SM, et al. Evaluation of the healthy eating index-2015. J Acad Nutr Diet 2018;118(9):1622–33.
36. Johnson CL, Dohrmann SM, Burt VL, et al. National health and nutrition examination survey: Sample design, 2011-2014. Vital Health Stat 2014;2(162):1–33.
37. Liu F, Wanigatunga AA, Schrack JA. Assessment of physical activity in adults using wrist accelerometers. Epidemiologic Rev 2021;43(1):65–93.
38. Subar AF, Thompson FE, Kipnis V, et al. Comparative validation of the block, willett, and national cancer institute food frequency questionnaires : The eating at America's table study. Am J Epidemiol 2001;154(12):1089–99.
39. Bedogni G, Bellentani S, Miglioli L, et al. The fatty liver index: A simple and accurate predictor of hepatic steatosis in the general population. BMC Gastroenterol 2006;6(1):33.
40. Fedchuk L, Nascimbeni F, Pais R, et al. Performance and limitations of steatosis biomarkers in patients with nonalcoholic fatty liver disease. Aliment Pharmacol Ther 2014;40(10):1209–22.
41. Koehler EM, Schouten JN, Hansen BE, et al. External validation of the fatty liver index for identifying nonalcoholic fatty liver disease in a population-based study. Clin Gastroenterol Hepatol 2013;11(9):1201–4.
42. Calori G, Lattuada G, Ragogna F, et al. Fatty liver index and mortality: The Cremona study in the 15th year of follow‐up. Hepatology 2011;54(1):145–52.
43. Zelber-Sagi S, Webb M, Assy N, et al. Comparison of fatty liver index with noninvasive methods for steatosis detection and quantification. World J Gastroenterol 2013;19(1):57.
44. Liu F, Wanigatunga AA, Schrack JA. Assessment of physical activity in adults using wrist accelerometers. Epidemiol Rev 2021;43(1):65–93.
45. Wolff-Hughes DL, McClain JJ, Dodd KW, et al. Number of accelerometer monitoring days needed for stable group-level estimates of activity. Physiol Meas 2016;37(9):1447–55.
46. Trost SG, McIver KL, Pate RR. Conducting accelerometer-based activity assessments in field-based research. Med Sci Sports Exerc 2005;37(11):S531–43.
47. Chen T, Parker J, Clark J, et al. National Health and Nutrition Examination Survey: Estimation procedures, 2011–2014. Series: Vital and health statistics, series 2, data evaluation and methods research; no. 177; DHHS pub; no. 2018-1377
48. Klein RJ, Schoenborn CA. Age adjustment using the 2000 projected U.S. population. Healthy People 2010 Stat Notes 2001(20):1–10.
49. Haneuse S, VanderWeele TJ, Arterburn D. Using the E-value to assess the potential effect of unmeasured confounding in observational studies. JAMA 2019;321(6):602–3.
50. VanderWeele TJ, Ding P. Sensitivity analysis in observational research: Introducing the E-value. Ann Intern Med 2017;167(4):268–74.
51. Greenland S. Concepts and pitfalls in measuring and interpreting attributable fractions, prevented fractions, and causation probabilities. Ann Epidemiol 2015;25(3):155–61.
52. Newson RB. Attributable and unattributable risks and fractions and other scenario comparisons. STATA J 2013;13(4):672–98.
53. Concato J, Peduzzi P, Holford TR, et al. Importance of events per independent variable in proportional hazards analysis. I. Background, goals, and general strategy. J Clin Epidemiol 1995;48(12):1495–501.
54. von Elm E, Altman DG, Egger M, et al. The strengthening the reporting of observational studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. Epidemiology 2007;18(6):800–4.
55. Jakicic JM, Kraus WE, Powell KE, et al. Association between bout duration of physical activity and health: Systematic review. Med Sci Sports Exerc 2019;51(6):1213–9.
56. Friedenreich CM, Stone CR, Cheung WY, et al. Physical activity and mortality in cancer survivors: A systematic review and meta-analysis. JNCI Cancer Spectr 2020;4(1):pkz080.
57. Campbell KL, Winters-Stone KM, Wiskemann J, et al. Exercise guidelines for cancer survivors: Consensus statement from international multidisciplinary roundtable. Med Sci Sports Exerc 2019;51(11):2375–90.
58. Kim D, Murag S, Cholankeril G, et al. Physical activity, measured objectively, is associated with lower mortality in patients with nonalcoholic fatty liver disease. Clin Gastroenterol Hepatol 2021;19(6):1240–7.e5.
59. Fung TT, Rexrode KM, Mantzoros CS, et al. Mediterranean diet and incidence of and mortality from coronary heart disease and stroke in women. Circulation 2009;119(8):1093–100.
60. Hu EA, Steffen LM, Coresh J, et al. Adherence to the healthy eating index-2015 and other dietary patterns may reduce risk of cardiovascular disease, cardiovascular mortality, and all-cause mortality. J Nutr 2020;150(2):312–21.
61. Yoo ER, Kim D, Vazquez-Montesino LM, et al. Diet quality and its association with nonalcoholic fatty liver disease and all-cause and cause-specific mortality. Liver Int 2020;40(4):815–24.
62. Park SY, Kang M, Shvetsov YB, et al. Diet quality and all-cause and cancer-specific mortality in cancer survivors and non-cancer individuals: The multiethnic cohort study. Eur J Nutr 2022;61(2):925–33.
63. Huang J, Liao LM, Weinstein SJ, et al. Association between plant and animal protein intake and overall and cause-specific mortality. JAMA Intern Med 2020;180(9):1173–84.
64. Chalvon-Demersay T, Azzout-Marniche D, Arfsten J, et al. A systematic review of the effects of plant compared with animal protein sources on features of metabolic syndrome. J Nutr 2017;147(3):281–92.
65. Tasevska N, Park Y, Jiao L, et al. Sugars and risk of mortality in the NIH-AARP diet and health study. Am J Clin Nutr 2014;99(5):1077–88.
66. Malik VS, Li Y, Pan A, et al. Long-term consumption of sugar-sweetened and artificially sweetened beverages and risk of mortality in US adults. Circulation 2019;139(18):2113–25.
67. Ramne S, Alves Dias J, González-Padilla E, et al. Association between added sugar intake and mortality is nonlinear and dependent on sugar source in 2 Swedish population–based prospective cohorts. Am J Clin Nutr 2019;109:411–23.