A weight regain of 1.5 kg or more and lack of exercise are associated with nonalcoholic fatty liver disease recurrence in men

The importance of maintaining the remission of nonalcoholic fatty liver disease (NAFLD) has been overlooked. Here we aimed to clarify factors causing NAFLD recurrence. In this retrospective cohort study over 10.8 ± 5.4 years, we investigated 1260 male health check-up participants diagnosed with NAFLD who achieved remission. The data were compared between the maintained remission and recurrence group. Among all participants, 618 (49.0%) showed NAFLD recurrence at the last visit. Participants in the maintained remission group continued to lose weight (72.7 ± 9.1, 68.7 ± 8.5 and 68.2 ± 8.9 kg), whereas those in the recurrence group lost and regained weight (72.9 ± 9.9, 69.7 ± 9.3 and 73.0 ± 10.4 kg). Receiver operating characteristic curve analysis showed a weight regain of + 1.5 kg as the cutoff value for recurrence. The proportion of regular exercisers at the last visit was 34.6% in the maintained remission group and 24.5% in the recurrence group (p < 0.0001). Multivariable analysis revealed the amount of weight regain (in 1 kg increments; adjusted odds ratio, 1.29; 95% confidence interval, 1.24–1.34) and regular exercise at the last visit (adjusted odds ratio, 0.67; 95% confidence interval, 0.55–0.89) were independently associated with recurrence. These findings demonstrate a weight regain of 1.5 kg or more and lack of exercise were associated with NAFLD recurrence.

The number of patients with nonalcoholic fatty liver disease (NAFLD) is increasing worldwide 1 . Some individuals with NAFLD develop nonalcoholic steatohepatitis (NASH), which causes liver fibrosis and increases their risk of hepatocellular carcinoma 2,3 . NAFLD has mutual and bi-directional association with metabolic syndrome, which includes hypertriglyceridemia, low high-density lipoprotein (HDL) level, and glucose intolerance [4][5][6] . Further, patients with NAFLD are at an increased risk of developing cardiovascular disease 7,8 . In addition, NAFLD and type 2 diabetes mellitus form a vicious cycle in the progression of both conditions 9,10 . The development of strategies for controlling NAFLD is required to prevent the progression of type 2 diabetes 11 .
NAFLD is reversible to a certain extent. Various reports have shown that weight loss and lifestyle interventions effectively improve NAFLD 12,13 and NASH 14,15 . The Japan Society of Hepatology recommends a body weight loss of > 7% to improve NAFLD and NASH. With respect to lifestyle modifications, exercise and calorie restriction are crucial for NAFLD improvement 16,17 . However, thus far, most studies have focused on treatments to achieve NAFLD remission. Although a fair number of studies examined the factors contributing to NAFLD recurrence after liver transplantation 18,19 , very few studies have investigated the natural history of patients who achieved remission through lifestyle changes. Thus, the factors that determine whether individuals maintain NAFLD remission or develop recurrence remain unclear.
To answer this question, we designed this retrospective cohort analysis of health check-up participants and examined changes in lifestyle history such as exercise and smoking habits obtained by questionnaires as well as clinical and laboratory data including metabolic syndrome parameters. This study aimed to reveal the factors associated to NAFLD recurrence diagnosed by abdominal ultrasonography. Considering that sex differences do exist in the prevalence and risk factors of NAFLD 20 , we analyzed the data separately by sex.

Results
We conducted a cohort analysis of participants who underwent a medical health check-up including abdominal ultrasonography. The evaluations were performed in the Medical Health Checkup Center at Asahi University Hospital, Gifu, Japan. We named the longitudinal cohort analysis using this database NAGALA (NAFLD in Gifu Area Longitudinal Analysis).
The present cohort study of men is shown in the flow diagram in Fig. 1. From the men who were registered as NAGALA between May 1994 and Dec 2018, we recruited those in whom fatty liver was diagnosed using abdominal ultrasonography at the first visit and who underwent at least three health examinations during the observation period. We excluded participants with an alcohol consumption of ≥ 30 g/day (695 participants) and those who had known liver disease including viral hepatitis (73 participants). Among the remaining 3624 participants, we sorted out the participants who showed disappearance of fatty liver on abdominal ultrasonography before the last visit. The earliest time point when fatty liver disappeared was set as the time of NAFLD remission. Finally, 1260 participants were investigated in the male study.
The data were compared between the maintained NAFLD remission group, in which fatty liver disappeared and was not identified at the last visit; and the NAFLD recurrence group, in which fatty liver once disappeared but reappeared at the last visit. Among these 1260 participants who achieved NAFLD remission during the observation period, 642 (51.0%) remained in remission until the last visit and 618 (49.0%) showed recurrence at the last visit (Fig. 1). We examined the data at three different time points: the first visit, time when fatty liver disappeared (NAFLD remission), and date of last visit.
The mean follow-up duration was 10.8 ± 5.4 years. The follow-up duration in the maintained NAFLD remission group was 10.0 ± 5.6 years and 11.6 ± 5.1 years in the NAFLD recurrence group (p < 0.0001).
The clinical and laboratory characteristics at baseline of the whole cohort and each group in men are summarized in Table 1. At baseline, the NAFLD recurrence group had a lower mean age, higher proportion of current smokers, higher mean serum aspartate aminotransferase level, higher mean serum alanine aminotransferase level, lower mean hemoglobin A1c (HbA1c) level, and higher mean serum triglyceride level than the maintained NAFLD remission group. There were no significant intergroup differences in mean body weight, body mass index (BMI), and exercise habits at baseline.
The mean body weight change at the three time points during the observation period is shown in Fig. 2. The participants in the maintained NAFLD remission group continued to lose weight (72.7 ± 9.1, 68.7 ± 8.5, and 68.2 ± 8.9 kg), whereas those in the NAFLD recurrence group lost and subsequently regained weight (72.9 ± 9.9, 69.7 ± 9.3, and 73.0 ± 10.4 kg). Two-way repeated-measures analysis of variance revealed a significant interaction of time and groups on body weight changes (p < 0.0001) (Fig. 2a). Between the first visit and the date of NAFLD remission, the mean weight reduction was -4.0 ± 4.5 kg in the maintained NAFLD remission group Participants whose fatty liver disappeared before the last visit Study population N = 1,260 Participants without fatty liver at the last visit Participants with fatty liver at the last visit   (Fig. 2b). Between the date of NAFLD remission and the last visit, however, a further body weight reduction of -0.5 ± 3.9 kg was observed in the maintained NAFLD remission group, while a weight regain of + 3.2 ± 4.6 kg was observed in the NAFLD recurrence group (p < 0.0001) (Fig. 2c).   Fig. 3a. The proportion of regular exercisers in the maintained NAFLD remission group continued to increase, whereas that in the NAFLD recurrence group increased between the first visit and the date of NAFLD remission and decreased between the date of NAFLD remission and the last visit. The proportion of regular exercisers at the last visit was 34.6% and 24.5% in the maintained remission and recurrence groups, respectively (p < 0.0001). No significant difference was observed in the proportion of participants who had newly started exercising by the date of NAFLD remission (19.9% vs. 19.0%, p = 0.6899) (Fig. 3b). After the date of NAFLD remission, however, the NAFLD recurrence group had a significantly lower proportion of participants who newly started exercising (15.4% vs. 9.2%, p = 0.0009) (Fig. 3c) and a higher proportion of participants who stopped regular exercise (9.8% vs. 12.3%, p = 0.1638) (Fig. 3d). The initiation or continuation of regular exercise at least once a week was associated with a lower incidence of NAFLD recurrence.
The proportion of current smokers continued to decrease in both groups. No significant intergroup difference was observed in the proportion of participants who had quit smoking by the date of NAFLD remission (6.8% vs. 7.1%, p = 0.8203) However, the proportion of participants who quit smoking was higher in the NAFLD recurrence group than in the maintained NAFLD remission group after NAFLD remission was achieved (5.6% vs. 8.6%, p = 0.0398). No significant intergroup differences were observed in alcohol consumption, coffee intake, and sleep duration.
The factors associated with NAFLD recurrence were assessed by univariable and multivariable logistic regression ( Table 2). The covariables included the baseline characteristics, body weight change, lifestyle changes, and p=0.6899 p=0.0009 p=0.1638  We also assessed the changes in metabolic syndrome parameters. The mean waist circumference at the last visit were 82.2 ± 6.8 cm and 86.3 ± 7.2 cm in the maintained remission and recurrence groups, respectively. The mean fasting blood glucose level remained low (5.98 ± 1.25 mmol/L) in the maintained NAFLD remission group and was higher (6.16 ± 1.48 mmol/L) in the recurrence group. The values of waist circumference, fasting blood glucose, HbA1c, systolic blood pressure, diastolic blood pressure, and serum triglycerides showed greater increases after remission in the recurrence group than in the maintained remission group. In addition, the mean serum HDL cholesterol level at the last visit was lower in the recurrence group than in the maintained remission group. Two-way repeated-measures analysis of variance revealed a significant interaction between time and groups on metabolic syndrome parameters (Fig. 4). This result indicated that NAFLD recurrence occurred simultaneously with worsening of metabolic syndrome parameters.
We compared metabolic syndrome parameters at the last visit and amount of weight change after NAFLD remission between the non-exercise and exercise groups of men (Table 3). Mean waist circumference was lower in the exercise group than the non-exercise group at the last visit (84.9 ± 7.2 cm vs. 82.7 ± 7.2 cm, p < 0.0001). The exercise group had lower serum triglyceride level (p = 0.0013) and higher HDL cholesterol level (p = 0.0002)    www.nature.com/scientificreports/ at the last visit. The amount of weight change after NAFLD remission was lower in the exercise group than the non-exercise group (+ 1.6 ± 4.6 kg vs. + 0.6 ± 4.7 kg, p = 0.0006). A receiver operating characteristic (ROC) curve analysis demonstrated that the weight change after NAFLD remission was able to predict the development of NAFLD recurrence reasonably well (area under the ROC curve = 0.75, p < 0.0001) (Fig. 5). The cutoff value of weight change that best identified patients with NAFLD recurrence was + 1.5 kg (sensitivity, 0.64; specificity, 0.73).
The cohort study in women was designed as shown in the flow diagram in Supplementary Fig. S1. Among 20,671 women who were registered in NAGALA between May 1994 and Dec 2018, the number of those with fatty liver at the first visit was 1878. We excluded participants whose alcohol consumption was ≥ 20 g/day and those who had a known liver disease including viral hepatitis. We sorted out the participants whose fatty liver disappeared before the last visit. Finally, 264 participants were included in the female study. Among the 264 female participants who achieved NAFLD remission during the observation period, 143 (54.2%) remained in remission until the last visit and 121 (45.8%) showed recurrence at the last visit ( Supplementary Fig. S1).
The clinical and laboratory characteristics at baseline in women are summarized in Supplementary Table S1. At baseline, the NAFLD recurrence group had a lower mean age, lower proportion of Fibrosis-4 (FIB-4) index (> 1.30 and < 2.67). The factors associated with NAFLD recurrence were assessed by univariable and multivariable logistic regression (Supplementary Table S2). The covariables included the characteristics at baseline that showed a significant intergroup difference: body weight change, follow-up period, age > 50 years (surrogate of menopausal state) at the last visit, exercise habits at the last visit, and quitting smoking after NAFLD remission. Multivariable logistic regression analysis showed that amount of weight change after the date of NAFLD remission (in 1 kg increments; aOR, 1.25; 95% CI, 1.15-1.35; p < 0.0001) and age > 50 years at the last visit (aOR, 4.01; 95% CI, 1.35-11.93; p = 0.0126) were independently associated with NAFLD recurrence.

Discussion
In the present study, 49.0% of the male participants experienced recurrence after NAFLD remission. This result indicated that many patients develop NAFLD recurrence despite achieving remission once through lifestyle changes. However, as the importance of maintaining NAFLD remission has been rather overlooked, little is known about the crucial factors related to NAFLD recurrence. This study provides new insight into the management of NAFLD to prevent relapse and its associated complications.
Of note, there were no significant differences between the maintained NAFLD remission and recurrence group in body weight or BMI at baseline (Table 1). Our study found that weight regain after NAFLD remission was strongly and independently associated with NAFLD recurrence (Fig. 2, Table 2). The area under the ROC curve represents the predictive ability of weight change for NAFLD recurrence (Fig. 5). The closer the value is to 1.0, the higher its predictive power. The area under the ROC curve was 0.75, which suggests that the predictive power of weight change had moderate accuracy. A weight change of + 1.5 kg after NAFLD remission (sensitivity, 0.64; specificity, 0.73) well identified patients with NAFLD recurrence. It suggested that a weight regain of + 1.5 kg or more would lead NAFLD recurrence and be a good indicator to avoid body weight rebound and maintain NAFLD remission.
As previous reports showed the impact of weight gain in the development of NAFLD 21,22 , our result indicated that weight regain was related to its recurrence. Although body composition was not evaluated in this study, we examined the correlation between the change in body weight and the change in waist circumference after NAFLD remission. They were positively well correlated (R 2 = 0.7005, p < 0.0001). This suggested that weight regain was associated with increasing amounts of visceral fat. NAFLD is defined by the accumulation of triglycerides in the liver. The sources of fatty acids that are required for the generation of hepatic triglycerides are principally derived from adipose tissue 23 . With increasing adipose tissue mass, specifically with the accumulation of dysfunctional and insulin-resistant adipocytes, release of fatty acids to the liver provides the substrate for and stimulus to hepatic lipid accumulation 24 .
Weight regain after successful weight loss is a major problem for many individuals. On average, more than half of the lost weight is regained within 2 years 25 . A previous study implied that weight loss-induced variations in cellular stress, extracellular matrix remodeling, inflammatory responses, adipokine secretion, and lipolysis are associated with the amount of weight that is regained after a successful weight loss 26 . Our results also demonstrated that regular exercise at least once a week is independently important to prevent NAFLD recurrence (Fig. 3, Table 2). The comparison of exercise and non-exercise groups in the present study revealed that exercisers regained less weight and showed smaller mean waist circumference, lower serum triglyceride level and higher serum HDL cholesterol level at the last visit (Table 3). Previous studies demonstrated clear evidence of the benefits of exercise therapy for reducing liver fat 27 and that the initiation of exercise is independently associated with NAFLD remission in men 22 . Consistent with the impact of exercise on NAFLD remission reported by previous studies, we revealed that the initiation or continuation of regular exercise is also important to preventing NAFLD recurrence. The mechanism for the direct hepatic benefit of regular exercise is inferred from rodent data. Regular exercise in rodents attenuates hepatic steatosis by reducing the lipogenic enzymes acetyl-coenzyme A carboxylase and fatty acid synthase 28 . Conversely, sedentary behavior in rats is marked by rapid reductions in hepatic fatty acid oxidation and mitochondrial enzyme activity (citrate synthase, β-hydroxyacyl-coA dehydrogenase, and cytochrome c oxidase), upregulation of fatty acid synthesis (increases in fatty acid synthase, stearoyl-coA desaturase-1, and sterol regulatory element binding protein 1c) 29 and elevation of malonyl-CoA, which inhibits fatty acid oxidation 30 .
The present study's findings indicated that quitting smoking is associated with NAFLD recurrence (Table 2). Although smoking cessation is known to decrease the risk of cardiovascular events 31 and decrease the all-cause mortality rate 32 , quitting smoking carries a certain risk for weight gain 33 , which is associated with NAFLD recurrence. Hu et al. demonstrated that smoking cessation accompanied by weight gain is associated with an increased short-term risk of type 2 diabetes but does not mitigate the benefits in terms of reducing cardiovascular and all-cause mortality 34 . Smoking cessation should be recommended to all smokers to reduce mortality from all causes; however, the risk of weight gain and NAFLD recurrence after cessation should be considered.
Metabolic syndrome includes visceral obesity, insulin resistance, alterations in glucose and lipid metabolism, and increased blood pressure 35 . By investigating the changes in metabolic syndrome parameters 36 in each group, we demonstrated that the values of waist circumference, fasting blood glucose, HbA1c, systolic blood pressure, diastolic blood pressure, and serum triglycerides were more elevated, while the mean HDL cholesterol level was lower in the NAFLD recurrence group than in the maintained remission group at the last visit (Fig. 4). Our findings are in line with the physiopathology of NAFLD, which has a mutual and bi-directional association with metabolic syndrome 4 . Therapies to prevent NAFLD recurrence are considered to reduce the risk of metabolic syndrome.
We also investigated the factors associated with NAFLD recurrence in women (Supplementary Tables S1,  S2). Of all female patients, 9.1% were diagnosed with NAFLD at the first visit ( Supplementary Fig. S1), a much lower proportion than that in the men (28.6% of all male participants). This finding was compatible with the fact that previous studies consistently indicated that the overall prevalence of NAFLD was higher in men than in women 37,38 . We also performed multivariate analysis including the covariable of age > 50 years as menopausal state at the last visit. Of note, the result showed that an age > 50 years at the last visit was independently associated with NAFLD recurrence (aOR, 4.01; 95% CI, 1.35-11.93; p = 0.0126), a finding in line with previous reports that mentioned NAFLD occurs at a higher rate in women after menopause, suggesting that estrogen is protective 20 . Although the analyses in women included a small population, we found that a weight change after remission was significantly related to NAFLD recurrence as in the male study. However, exercise habits at the last visit were not a significant factor in women.
Our study had several limitations. First, we categorized participants who regularly engaged in any kind of sport activity at least once a week as regular exercisers, whereas the exercise recommendation for health is generally 150 min of moderate-intensity physical activities plus resistance exercise twice a week by many organizations including the World Health Organization. Thus, it is important to evaluate physical activity mode, frequency, duration, and intensity. In a previous study, the hazard ratio for all-cause mortality was 0.66 (95% CI, 0.62-0.72) in insufficiently active participants who reported one to two exercise sessions per week versus inactive participants 39 . This suggests that any exercise performed more than once a week, which we defined as regular exercise in this study, contributes to health improvement.
Second, we did not assess the participants' dietary habits. Weight gain is strongly affected by food intake variations 40 . It has also been suggested that having a tendency toward disinhibition or food addiction is associated with weight regain after a weight loss intervention 41 . Therefore, a survey of dietary and eating behaviors is needed to elucidate the independent role of each lifestyle modification.
Third, we performed a qualitative evaluation by ultrasonography to diagnose NAFLD but no quantitative evaluations. Evaluating semi-quantitative ultrasonographic indices is a useful and non-invasive way to investigate NAFLD severity, which varied among the participants 42 .
Next, there was an intergroup difference in the observation duration. However, we emphasized evaluating changes in lifestyle history, clinical parameters, and laboratory data between the time points. We also revealed that the intergroup differences in body weight changes remained significant after the degrees of change were divided by each observation duration. The mean amount of weight change between the date of NAFLD remission and the last visit divided by the duration in years was − 0.2 ± 1.7 and + 0.7 ± 1.1 kg/year (p < 0.0001) in the maintained NAFLD remission and recurrence groups, respectively.
A weight regain of 1.5 kg or more and lack of exercise are associated with NAFLD recurrence and worsening metabolic syndrome parameters in men. This study's findings indicate the importance of weight loss maintenance and regular exercise for sustaining NAFLD remission.  44 . We used a high cutoff value of FIB-4 index ≥ 2.67 and a low cutoff value of FIB-4 index ≤ 1.30 45 . The diagnosis of fatty liver was made on the basis on the results of abdominal ultrasonography performed by trained technicians. All ultrasonographic images were stored as copies, and one gastroenterologist reviewed the images and made the diagnosis of fatty liver without referring to any other data. Of four known criteria (hepatorenal echo contrast, liver brightness, deep attenuation, and vascular blurring), the presence of both hepatorenal contrast and liver brightness was required to make a diagnosis of fatty liver 46 . To diagnose NAFLD, we excluded participants whose alcohol consumption was ≥ 30 g/day (≥ 210 g/week) for men and ≥ 20 g/day (≥ 140 g/week) for women 47 . We excluded known liver disease of drug-induced liver injury, cirrhosis, and fulminant hepatic failure according to what the participants stated in a questionnaire administered at the first visit. In addition, hepatitis B surface antigen and hepatitis C virus antibody was evaluated in the blood examination and we excluded the subjects when the result of either was positive at the first visit 48 .
Standardized questionnaire about lifestyle factors. A standardized questionnaire was administered to all participants. Alcohol consumption habits were evaluated according to alcohol drink amount and type consumed per week during the past month and the mean ethanol intake per week was estimated. The participants were also categorized into three groups according to smoking status (never, former, and current). Exercise habits were asked in the questionnaires as follows: (1) Do you exercise regularly, occasionally, or not at all? and (2) How many times per week do you exercise? When the participants answered that they exercised regularly once or more a week, we categorized them into those who exercised regularly.

Statistical analysis.
Continuous values are presented as mean ± SD or median (interquartile range), while categorical values are presented as n (%). Student's t test or the Wilcoxon signed-rank test were used to analyze the statistical differences in continuous variables. The analysis of categorical variables between the groups was performed using Pearson's χ 2 test. Two-way repeated-measures analysis of variance was performed to determine the existence of a significant interaction between time and group. We used univariable and multivariable logistic regression analyses to assess factors associated with NAFLD recurrence. An ROC curve analysis was conducted to evaluate the predictive ability of weight change for NAFLD recurrence and identify cutoff values of weight change that best identified participants with NAFLD recurrence. The cutoff value was at a point with high sensitivity and high specificity. All statistical analyses were performed using JMP Pro 15 software (https:// www. jmp. com/ en_ us/ softw are/ predi ctive-analy tics-softw are. html).