Risk factors for development of nonalcoholic fatty liver disease after pancreatoduodenectomy

Abstract Considerable attention has been focused on nonalcoholic fatty liver disease (NAFLD) which occasionally develops after pancreatoduodenectomy (PD). The present study aimed to clarify the prevalence, sequential change in properties and risk factors for NAFLD development after PD. We enrolled 196 patients who underwent PD and a computed tomography (CT) scan 1 month, 6 months and 1 year after surgery. NAFLD was defined as a liver‐to‐spleen attenuation ratio on plain CT of <0.9. We compared various clinical factors between the NAFLD group and the control group. Individual prevalence of NAFLD at 1 month, 6 months and 1 year after surgery was 12%, 21% and 15%. Significantly different factors by univariate analysis were as follows: 1 month: age, sex, total protein (TP), total cholesterol (TC) and copper (Cu); 6 months: sex, disease, surgical method, portal vein resection (PVR), frequency of defecation, TC and Cu; 1 year: age, sex, disease, surgical method, PVR, frequency of defecation, TP and Cu. Risk factors by multivariate analysis were as follows: 1 month: not elderly age, female sex and a decrease in Cu; 6 months: female sex and a decrease in Cu; 1 year: a decrease in Cu. NAFLD after PD frequently developed in women with a decrease in serum Cu and was influenced by various factors related to poor digestion and absorption associated with pancreatic exocrine insufficiency.

Etiology of this steatosis may be different from that of conventional nonalcoholic fatty liver disease (NAFLD) related to obesity, dyslipidemia or diabetes mellitus (DM). Post-PD NAFLD is believed to be most commonly a result of insufficiency of digestion, absorption, nutrition and pancreatic exocrine function, but clarification of this etiology remains an important unsolved problem. Moreover, risk factors for NAFLD at specific points in time after PD have yet to be identified, so elucidation of these factors will contribute to the prevention and management of this complication. To investigate the prevalence and risk factors for NAFLD after PD, we retrospectively studied patients undergoing PD in our hospital. were enrolled in the present study, and they underwent physical, radiological and blood examinations at 1 month, 6 months and 1 year after PD. No patient had regular alcohol intake over 140 g/week after surgery, who was excluded from this study. Informed consent for this retrospective analysis was obtained by the opt-out method, and this study was approved by the institutional review board at Miyazaki University in 2016 (approval number #2016-081).

| Definition of NAFLD
CT images were obtained with a 128-row multidetector CT scanner (Siemens, Munich, Germany) using 120 kVp, 0.7 pitch, 5 mm slice thickness and 0.5 s scan time. NAFLD was defined as a liver-tospleen attenuation ratio of <0.9 2,3,14 or a hepatic attenuation value at least 10 Hounsfield units lower than the spleen attenuation value 2,15 on unenhanced abdominal CT scan. We used average CT attenuation values in four sectors of the liver and the median value from all sections of the spleen (Figure 1). Prevalence of NAFLD was evaluated at 1 month, 6 months and 1 year after PD.

| Pancreatic enzymes
Pancreatic enzyme supplements such as Toughmac â 3 g/day (Ono Pharmaceutical Co, Ltd, Osaka, Japan) or Berizym â 3 g/day (Shionogi & Co. Ltd, Osaka, Japan) were routinely given in 124 patients. Since 2012, we have used the pancrelipase delayed-release supplement LipaCreon â 1200 mg/day (Eisai Co. Ltd, Tokyo, Japan) in 59 patients. At the attending physician's discretion, the remaining 13 patients did not receive any type of pancreatic enzyme supplement.
Thus, to analyze the difference in the rate of NAFLD development according to the types of drug supplement given, we classified the patients into the following three groups: none (n=13), conventional drugs (n=124) and pancrelipase (n=59).

| Clinical parameters
To investigate the clinical parameters associated with an occurrence of NAFLD after PD, we compared the following data between the patients with and without NAFLD at three time points: preoperatively: age, sex, body mass index (BMI), type of disease (PDC/other diseases), DM and dyslipidemia; perioperatively: surgical method (PPPD/SSPPD/PD), combined surgery, PVR and frequency of defecation at discharge from initial hospital stay; and postoperatively: types of pancreatic enzyme supplement and blood tests. Blood tests included serum values of total protein (TP, g/dL), total cholesterol F I G U R E 1 We measured the average computed tomography (CT) attenuation values in four sectors of the liver and the median value from all sections of the spleen on unenhanced CT scans. Nonalcoholic fatty liver disease was defined as a liver-to-spleen attenuation ratio of <0.9 or a hepatic attenuation value at least 10 Hounsfield units lower than the spleen attenuation value.

| Statistical analysis
Clinical parameters are expressed as the number (percentage) or median (range). Comparisons between groups were made using the v 2 test for categorical variables and the Mann-Whitney U-test for continuous variables. Multivariate logistic regression analysis by backward elimination method was also conducted to determine the risk factors associated with the development of NAFLD after PD. All P values were based on a two-sided test, and a P value <.05 was considered to be statistically significant. Statistical analyses were carried out using SPSS software version 18.0 for Windows (SPSS, Chicago, IL, USA).

| RESULTS
Nonalcoholic fatty liver disease developed before surgery in one patient only. This patient had advanced pancreatic cancer invading the duodenum, which caused duodenal obstruction and severe atrophy of the distal pancreas. NAFLD in this patient was improved at 1 month after PD and did not recur from 6 months after PD.
Development of NAFLD at 1 month, 6 months and 1 year after PD was detected in 23 (12%), 41 (21%) and 30 (15%) patients, respectively. Of the 41 patients in whom NAFLD developed 6 months after PD, 18 recovered spontaneously by 1 year after, and 25 had not suffered from NAFLD 1 month after PD. Sequential change in patients with NAFLD at 1 month, 6 months and 1 year after PD is summarized in Figure 2.
Comparison of laboratory data between the patients with and without NAFLD by univariate analysis showed a significant decrease in age, BMI, levels of serum TP, TC and Cu 1 month after PD in the NAFLD group. Additionally, by univariate analysis, sex and rate of DM were significantly different between the two groups (Table 1).
At 6 months after PD, sex, type of disease, surgical method and PVR rate were significantly different between the two groups by univariate analysis. Frequency of defecation was increased and levels of serum TC and Cu were decreased significantly in the NAFLD group by univariate analysis (Table 2).
At 1 year after PD, age and levels of serum TP and Cu were decreased, and frequency of defecation was increased significantly in the NAFLD group by univariate analysis. Sex, type of disease, surgical method and PVR rate were also significantly different between the two groups by univariate analysis (Table 3).
Age (not elderly), female sex and a decrease in Cu were related to an increased risk of developing NAFLD 1 month after PD by multivariate analysis (Table 4). Multivariate analysis also showed female sex and a decrease in serum Cu 6 months after PD, and a decrease in serum Cu 1 year after PD to be associated with an increased risk of developing NAFLD (Table 4).

| DISCUSSION
Individual prevalence of NAFLD at 1 month, 6 months and 1 year after PD was 12%, 21% and 15%. These values seemed to be lower than those reported in previous studies 2-6,11,12 because the development of NAFLD was not evaluated at a fixed time in most of the previous studies. Indeed, we recognized that several patients recovered from NAFLD spontaneously and others deteriorated sequentially, even though they were all taking the same pancreatic enzyme supplements for 1 year (Figure 2). Which patients were likely to suffer or recover from NAFLD remains uncertain from our study, but this is a first report documenting the sequential change in NAFLD development after PD.
PD can cause a diverse group of diseases as a result of impaired gut motility, insufficiency of neuroendocrine hormones, or malabsorption of certain trace elements. Previous reports found pancreatic head cancer to be an independent risk factor for the development of post-PD NAFLD, 2,3 which was in accord with our results from 6 months after PD. Massive resection of the proximal pancreatic parenchyma is often necessary to obtain surgical curability of PDC, which results in insufficient exocrine function of the remnant pancreas. Nakagawa et al. 6 reported that postoperative pancreatic exocrine insufficiency measured by 13 C-labeled mixed triglyceride breath test was closely associated with NAFLD. Pancrelipase delayedrelease supplement (LipaCreon â 1200 mg/day) could improve exocrine dysfunction and was reported to have a significant beneficial F I G U R E 2 Patients with or without nonalcoholic fatty liver disease (NAFLD) were plotted as filled or open circles, respectively. NAFLD development was improved sequentially from 1 month to 6 months or 1 year after surgery in several patients. In contrast, NAFLD developed 6 months after surgery in others and was improved at 1 year after pancreatoduodenectomy in some patients. effect on NAFLD after PD. 7,13 However, the prevalence of NAFLD has not decreased since 2012 when we started giving pancrelipase supplements in our hospital. A certain percentage of patients suffered NAFLD even if they received pancrelipase supplements, so we believe that the advantage of this drug is equivocal for the prevention of NAFLD development although it is helpful for intestinal digestion and absorption.
The most notable finding of the present study was that female sex was one of the most serious risk factors for the development of post-PD NAFLD. The number of women was significantly higher in the NAFLD group than in the control group at all three time points after PD by univariate analysis. Multivariate analysis also showed that female sex was one of the risk factors at 1 month and 6 months after PD, so this appears to be quite an important factor associated with the development of post-PD NAFLD. Song et al. 5 reported that female sex was a significant risk factor but did not describe the reason for or emphasize this important finding in their report. Sato et al. 12 also showed evidence that female sex was one of the risk factors for the development of post-PD NAFLD by multivariate analysis as did the present study. They suggested the same mechanism in women as that of conventional NAFLD, which might be influenced by an imbalance in sex hormones after menopause in women over the age of 50. 16 We also presume that the change in body fat distribution after menopause 17 and the accumulation of visceral fat markedly accelerated by menopause 18 can contribute to the high incidence of post-PD NAFLD seen in women.
Not elderly age, which was one of the risk factors for NAFLD at 1 month and 1 year after PD by univariate analysis in this series, might also be associated with the timing of menopause.
Cu is a trace element that is essential to the function of several enzymes involved in a variety of biological processes. Cu deficiency is known to affect the cardiovascular, musculoskeletal and nervous systems, to impair the metabolism of cholesterol and glucose and to affect the function of the hepatic antioxidative system. 19 Aigner et al. 20 reported that serum Cu levels were lower in patients with conventional NAFLD than in controls. Ackerman et al. 21 suggested that Cu deficiency in rats was a link between a fructose-enriched diet and the development of NAFLD. They also reported that decreases in plasma and hepatic Cu concentrations correlated with a decrease in the hepatic expression of copper-zinc-superoxide dismutase, disappearance of hepatic metallothionein and an increase in the expression of hepatic nitrotyrosine. Moreover, they showed the  dietary Cu-deficient rats exposed to fructose. They suggested that high fructose-induced NAFLD might be due, in part, to inadequate dietary Cu and that impairment of duodenum Cu transporter-1 expression seen in fructose feeding might lead to decreased Cu absorption and subsequent Cu deficiency. In our study, a decrease in serum Cu was one of the most serious risk factors for NAFLD development at all three time points after PD by multivariate analysis.
Therefore, we speculate that Cu supplementation might have therapeutic potential for post-PD NAFLD ascribed to Cu deficiency.
Zn deficiency is one of the most consistent biochemical manifestations in alcoholic liver steatosis. Kang et al. 23 reported that Zn supplementation reversed alcohol-induced steatosis in mice, the mechanisms of which were by potentially inhibiting oxidative stress and reactivation of hepatocyte nuclear factor-4a and peroxisome proliferators activated receptor-a. However, from our results, Zn metabolism seemed not to contribute to the development of post- Although there is overwhelming evidence of an association between insulin resistance and the prevalence of conventional NAFLD, 24 the value of HOMA-R was not significantly different between the two groups and the prevalence of accompanying DM was paradoxically less in the NAFLD group than in the controls at 1 month after PD in this series. This is possibly because the probability of progressive stages of NAFLD, such as non-alcoholic steatohepatitis (NASH) and liver fibrosis increases with the presence of impaired glucose tolerance. 25 Hamaguchi et al. 26 advocated that tight glycemic control might prevent histological progression, such as liver fibrosis in Japanese patients with NAFLD. Insulin resistance is probably not related to the development of post-PD NAFLD itself but to the progression from NAFLD to NASH or more advanced stages.
Our results showed that a decrease in serum TP or TC level was one of the risk factors for NAFLD at any of the three time points after PD by univariate analysis. This factor indicates malnutrition of the patients, but whether this is a cause of NAFLD or a result of it was unclear. In this series, SSPPD was a significant risk factor from 6 months after surgery by univariate analysis. Nanashima et al. 27 reported that the incidence of NAFLD was significantly higher in the patients who received SSPPD than in those receiving PPPD. In the late period, SSPPD could worsen the nutritional state of the patients more than PPPD. 28 Elucidation of the molecular mechanisms may lead to an understanding of paradoxical hepatic fat accumulation in the malnourished state.
Little information about the influence of PVR on the development of post-PD NAFLD was obtained. In this series, PVR imparted an increased risk of NAFLD from 6 months after PD by univariate analysis. This factor might worsen the portal venous flow in the late period after PD, which we did not evaluate in the present study.
In conclusion, NAFLD after PD frequently developed in women with a decrease in serum Cu and was influenced by various factors related to malnutrition involving pancreatic exocrine insufficiency such as PDC, SSPPD, PVR and high frequency of defecation.
Because of the limitations of small sample size and the single-institution nature of this study, further prospective or multi-institutional joint research will be necessary to clarify the detailed mechanism of post-PD NAFLD.