Post-transplant obesity impacts long-term survival after liver transplantation

Because pre-transplant body mass index (BMI) can be confounded by ascites, we hypothesized that post- transplant BMI at 1 year could predict long-term survival. Methods: A post-hoc analysis was performed of an observational cohort study consisting of adult recipients of a ﬁ rst OLT between 1993 and 2010. Baseline BMI was measured at 1-year post-transplantation to represent a sta-blecondition.Recipientswerestrati ﬁ edintonormalweight(BMI b 25kg/m 2 ),overweight(25 ≤ BMI ≤ 30kg/m 2 ), and obese (BMI N 30 kg/m 2 ). Kaplan-Meier survival analyses were performed with log-rank testing, followed by multivariable Cox proportional hazards regression analysis. Results: Out of 370 included recipients, 184 had normal weight, 136 were overweight, and 50 were obese at 1- year post-transplantation. After median follow-up for 12.3 years, 107 recipients had died, of whom 46 (25%) had normal weight, 39 (29%) were overweight, and 22(44%) were obese (log-rank P = 0.020). Obese recipients hadasigni ﬁ cantlyincreasedmortalityriskcomparedtonormalweightrecipients(HR2.00,95%CI1.08 – 3.68, P = 0.027).BMIwasinverselyassociatedwith15years patientsurvival(HR1.08,95%CI1.03 – 1.14, P =0.001perkg/ m 2 ),independentofage,gender,musclemass,transplantcharacteristics,cardiovascular riskfactors,kidney-and liver function. Conclusion: Obesity at1-year post-transplantation conveysa 2-foldincreasedmortality risk,whichmay offer potential for interventional strategies (i.e. dietary advice, lifestyle modi ﬁ cation, or bariatric surgery) to improve long-term survival after OLT.


Introduction
Orthotopic liver transplantation (OLT) is the life-saving treatment for patients suffering from end-stage liver disease [1]. Although shortterm survival has improved over the past decades, long-term survival remains impaired. Overall 1-and 5-year survival rates after OLT are 90% and 70% respectively [2,3], whereas 20-year survival rate is approximately 50% [4]. This emphasizes the importance to identify risk factors for long-term outcomes in OLT recipients. Several risk factors have already been shown to impair long-term outcomes after OLT, such as primary liver disease [5,6], older recipient and donor age [5][6][7], as well as the livelong dependency on chronic immunosuppression therapy [8], which predisposes to the development of de novo malignancies [9], renal dysfunction [10,11], hypertension [2,12], new onset of diabetes mellitus [13,14], and hyperlipidaemia [12]. However, the effects of overweight and obesity on long-term survival after OLT remain contradictory [15].
Overweight and obesity are characterized by an abnormal or excessive fat accumulation that may impair health and is measured by the body mass index (BMI) [16]. In the general population, overweight and obesity are a major problem. Overall, in 2016, approximately 39% of the world's adult population was overweight, and 13% were obese [16]. Furthermore, the worldwide prevalence of obesity nearly tripled between 1975 and 2016. In the general population, high BMI is a risk factor for cardiovascular diseases, diabetes, musculoskeletal disorders, and cancer, resulting in increased morbidity and mortality [16]. Pre-transplantation overweight and obesity are very common in OLT recipients. The prevalence of overweight and obesity has been reported to be 41% and 28%, respectively [17]. Additionally, alcoholic liver disease, hepatocellular carcinoma, genetic factors, and male gender are risk factors for new-onset obesity after liver transplantation [18]. History of smoking and higher age are also associated with an increased incidence of obesity after OLT [19]. Because OLT recipients are prone to develop overweight or obesity post-transplantation [20], it is important to investigate the effects of overweight and obesity on long-term survival after OLT.
As pre-transplant body weight can be confounded by ascites, it is estimated that 11-20% of patients with large volume ascites are misclassified as being overweight or obese [21,22]. Furthermore, OLT recipients gain weight after transplantation, mainly during the first year after OLT [23]. Therefore, post-transplant body weight is likely to be more representative to calculate true BMI. In this study, we hypothesized that post-transplant BMI at 1 year after OLT is associated with long-term survival.

Study design and population
A post-hoc analysis of an observational cohort study (www. trialregister.nl -Trial NL6334) of adult (age ≥18 years) patients, who underwent a first OLT between 1993 and 2010, was performed. Baseline was set at 1 year post-transplantation to represent a stable condition and because most weight gain occurs within the first year after transplantation [23]. OLT recipients with missing baseline data on BMI, age, gender, or urinary creatinine excretion rate (CER) were excluded. Furthermore, those OLT recipients who died within 1 year after transplantation and those who were lost to follow-up were excluded. OLT recipients were stratified into normal weight (BMI b 25 kg/m 2 ), overweight (25 ≤ BMI ≤ 30 kg/m 2 ), and obese (BMI N 30 kg/m 2 ). This study was approved by the Medical Ethical Committee of our institute (METc 2014/77) and adhered to the Declaration of Helsinki and the Declaration of Istanbul.

Data collection
Gender, age, current smoking status, height, weight, primary liver disease, complications, and medication use were derived from patient records. Transplant characteristics were derived from the recipient's operative report. Donor characteristics were retrieved from the Eurotransplant database.
BMI was obtained by dividing a person's weight by the square of the person's height (kg/m 2 ). Body surface area (BSA) was calculated using the DuBois formula [24]. Cardiovascular disease history was defined as a previous myocardial infarction, cerebrovascular accident, and/or peripheral arterial disease. Hypertension was defined as a blood pressure of N140/90 mmHg and/or the use of antihypertensive medication.
Cumulative prednisolone dose was calculated as the daily prednisolone dose at baseline, multiplied by the number of days since transplantation, adding the dosage of prednisolone or methylprednisolone given for treatment of rejection. Methylprednisolone dosage was converted into prednisolone equivalents by multiplying methylprednisolone dosage by a factor of 1.25 [25]. To account for differences in immunosuppressive regimes, transplantation dates were stratified into 3 era's. During the first era (1993-1998), a combination of prednisolone Note to Table 1: Data are represented as mean ± SD, median (interquartile range) or n (%). Differences were tested by ANOVA or Kruskal-Wallis for continuous variables and with χ 2 -test for categorical variables. BMI, body mass index; BSA, body surface area; CER; creatinine excretion rate; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; AST, aspartate aminotransferase; ALT, alanine aminotransferase; γ-GT, gamma-glutamyltransferase; ALP, alkaline phosphatase; NASH, non-alcoholic steatohepatitis; ICU, intensive care unit; MELD, model for end-stage liver disease. (10 mg/day), azathioprine (125 mg/day), and cyclosporin A (dosage resulting in trough blood levels of 100 μg/L) was given. During the second era (1999-2004), a combination of prednisolone and tacrolimus (dosage resulting in trough blood levels between 5 and 7 μg/L) could be given, as well as the combination of prednisolone, azathioprine, and cyclosporin A. Finally, during the third era (2005-2010), a combination of prednisolone, mycophenolate mofetil, and tacrolimus was given.
All serum and urine laboratory parameters were derived from our center's electronic laboratory system, using the median values between 9-and 15-months post-transplantation to minimize collection and measurement errors. OLT recipients were instructed to collect their urine according to a standardized protocol. For 24 subsequent hours, recipients collected urine, excluding the morning urine of the first day and including their morning urine of the second day. Kidney function was determined using the estimated glomerular filtration rate (eGFR), which was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula [26]. Proteinuria was defined as urinary protein excretion of N0.5 g/day. Muscle mass was estimated using urinary CER, obtained from 24 h urine samples [25].
Causes of death were derived from patient records or requested from general practitioners, and categorized into cardiovascular, infectious, malignant, and miscellaneous.

Outcome measures and follow-up
The primary outcome of this study was 15 years all-cause mortality. The secondary outcome was cause-specific mortality divided into four categories: cardiovascular, infectious, malignancy, and miscellaneous. Follow-up was recorded up to 15 years after baseline, or until December 31, 2018.

Statistical analyses
Continuous variables are presented as mean ± standard deviation (SD) if normally distributed and as median (interquartile range [IQR]) if skewed. Categorical variables are presented as number (percentage). Differences across BMI stratified groups were compared using the oneway analysis of variance (ANOVA) for normally distributed variables, the Kruskal-Wallis test for skewed distributed variables, and chisquare test for categorical variables.
Kaplan-Meier analysis with the log-rank test was used for initial survival analysis. Subsequently, Schoenfeld residuals were investigated to test the proportionality of hazards. Cox proportional hazards regression analyses were performed for BMI as categorical variable as well as continuous variable. Data were presented as hazard ratios (HR) and 95% confidence intervals (CI). Potential interactions of BMI with age, gender,  and urinary CER were checked. Cox proportional hazards models were constructed to adjust for potential confounders.
In model 1, we performed a crude Cox proportional hazards regression analysis. Subsequently, multivariable Cox proportional hazards regression analyses were performed. In model 2, we adjusted for body composition, using age, gender, and muscle mass as measured urinary CER. In model 3, we cumulatively adjusted for transplant related factors, including primary liver disease and transplantation era. In model 4 we cumulatively adjusted for independent cardiovascular risk factors, including cardiovascular disease history and smoking status. In model 5, we cumulatively adjusted for kidney function, using eGFR and the presence of proteinuria. Finally, in model 6, cumulative adjustments were made for liver function, using liver enzymes (AST, ALT, gamma-GT, ALP), direct bilirubin, and serum albumin. No adjustments were made for the use of calcineurin inhibitors and prednisolone, since transplantation eras are based on medication regimes. However, variations in the standard regimens were present and were related to side effects or treatment of allograft rejection [27,28]. Therefore, sensitivity analyses were performed, replacing transplantation era by the use of calcineurin inhibitors and/or prednisolone and/or cumulative prednisolone dose.
Additionally, univariable and multivariable Cox proportional hazards regression analyses were performed for all baseline variables, excluding variables with N10% missing values. Multivariable analysis was performed including all variables with P b 0.1 in the univariate analysis and gender. When variables were represented by other variables (e.g. serum creatinine and eGFR), the most significant variable was included in the multivariable analysis.
Cause-specific mortality was assessed using cox proportional hazards analysis and subsequently, predictors of cardiovascular mortality were assessed, using competing-risks regression models according to Fine and Gray [29].
To assess the effect of change in BMI in the first-year posttransplantation compared to pre-transplantation on all-cause mortality, additional analyses were performed using the models described above. Change in BMI was calculated as BMI 1-year post-transplantation minus BMI pre-transplantation. For BMI pre-transplantation the last documented BMI before transplantation was used. Because pre-transplant BMI is affected by ascites, we additionally performed analyses correcting for pre-transplant ascites in model 2, and excluding patients with pre-transplant ascites.
To visualize the relationship between BMI and all-cause mortality, a multivariable adjusted, restricted cubic spline with 3 knots positioned at the 10th, 50th, and 90th was made, based on model 6. Median BMI was used as reference.
P values are two-tailed and for all analyses a P value of b0.05 was considered to be statistically significant. Statistical analyses were performed, using IBM Statistics SPSS version 23.0 (IBM Inc. Chicago, IL), Stata/SE 14.2 (StataCorp LLC. College Station, TX), GraphPad Prism 7.02 (GraphPad Software, Inc. San Diego, CA), and R version 3.5.1 (R Foundation, Vienna, Austria).

Results
Between 1993 and 2010, a total of 393 adult patients underwent OLT. Nine OLT recipients died within the first year after transplantation and were excluded, as well as 13 OLT recipients with missing baseline data on BMI. Additionally, one recipient was lost to follow-up, resulting in 370 OLT recipients eligible for analysis. The majority of 184 (49.7%) OLT recipients had a normal weight, whereas 136 (36.8%) were overweight, and 50 (13.5%) were obese.
Baseline characteristics according to BMI categories are described in Table 1. Of all OLT recipients, 214 (57.8%) were male, with no significant differences between groups. Mean age was 48.3 ± 12.5 years, with normal-weight OLT recipients being significantly younger than those with a higher BMI classification. Mean BMI was 25.7 ± 4.5 kg/m 2 . Concerning body composition, weight, BMI, and BSA, were significantly higher, whereas height was significantly lower in the obese group compared to lower BMI groups. However, no significant differences were found in urinary CER between groups. Hypertension was present in 225 (60.8%) OLT recipients and significantly higher in the obese group compared to lower BMI groups. Mean eGFR was 69.7 ± 21.6 ml/min/ 1.73m 2 , with normal-weight OLT recipients having a significantly higher eGFR than overweight or obese OLT recipients. Serum triglycerides and glucose were significantly higher in the obese group compared to lower BMI groups, as well as the use of antidiabetics, antihypertensives, and statins.
Significant differences were seen between BMI groups and primary liver disease (P b 0.001). Primary sclerosing cholangitis (PSC) was most common in normal weight OLT recipients, viral hepatitis and alcohol cirrhosis were most common in overweight OLT recipients, and cryptogenic cirrhosis/non-alcoholic steatohepatitis (NASH) and alcohol cirrhosis were most common in obese OLT recipients.
Kaplan-Meier survival curves are depicted in Fig. 1. During a median follow-up of 12.3 years (IQR 8.4-15.0 years), 107 (28.9%) OLT recipients  (Fig. 1, log-rank test: P = 0.020; Fig. 2). Twenty-four (22.4%) OLT recipients died as a result of cardiovascular causes, 28 (26.2%) OLT recipients died due to infectious causes, and for 33 (30.8%) OLT recipients their cause of death was malignant disease. Additionally, 22 (20.6%) OLT recipients died as a result of other causes, including recurring liver cirrhosis (n = 8), amyloidosis (n = 6), suicide or euthanasia (n = 3), neurologic causes (n = 2), graft failure (n = 1), lung emphysema (n = 1), and kidney failure due to azathioprine use (n = 1). No significant differences were found in causes of death between BMI groups ( Table 2). Tables 3 and 4. We found no evidence for interactions of BMI with age, gender, or urinary CER (all P ≥ 0.05). Analyses according to BMI categories showed that obese OLT recipients had a significantly 2-fold higher risk of all-cause mortality (Table 3, model 1: HR = 2.04, 95% CI: 1.23-3.39, P = 0.006), when compared to normal weight OLT recipients. Multivariable adjustments for age, gender, urinary CER, primary liver disease, transplantation era, cardiovascular risk factors, kidney function, and liver function ( Table 3, models 2-6) did not materially alter these results (Table 3, model 6: HR = 2.00, 95% CI: 1.08-3.68, P = 0.027). As a continuous variable, BMI was associated with higher all-cause mortality (Table 3, model 1 and Table 4, HR = Univariable and multivariable Cox proportional hazards regression analyses were performed for all variables with b10% missing values. Multivariable analysis was performed including all variables with P b 0.1 in the univariate analysis and gender. When variables were represented by other variables (e.g. serum creatinine and eGFR) the most significant variable was included in the multivariable analysis.
Sensitivity analyses replacing transplantation era in model 3 by the use of calcineurin inhibitors and/or the use of prednisolone and/or cumulative prednisolone dose did not materially change the results for continuous and stratified analyses (Supplementary Table S1).
Additional Cox regression analyses were performed to investigate the association of BMI with cause-specific mortality ( Table 5). A significant association of BMI with infectious mortality was found (Table 5, model 6: HR = 1.12, 95% CI: 1.00-1.25, P = 0.046). We found no statistically significant associations for BMI with cardiovascular, malignant, and miscellaneous mortality. Competing-risks regression models on cardiovascular mortality are described in Supplementary Table S2 Furthermore, change in BMI in the first-year post-transplantation compared to pre-transplantation, was analysed (Table 6). BMI pretransplantation was measured at a median of 35.0 (IQR 8.5-80.0) days before transplantation. The vast majority of obese recipients 1-year posttransplantation were either overweight, or obese prior to transplantation. Normal-weight OLT recipients prior to transplantation rarely progressed to obesity after 1 year (Table 6). Change in BMI was not significantly associated with all-cause mortality (Supplementary Table S3). Results remained non-significant after correction for pre-transplant ascites (Supplementary Table S4). Analyses excluding OLT recipients with pretransplant ascites, showed similar results (Supplementary Table S5).

Discussion
This study demonstrates that high BMI at 1 year after OLT is associated with increased risk of long-term all-cause mortality in OLT recipients. Furthermore, obese OLT recipients have a 2-fold increased risk of long-term all-cause mortality compared to normal weight OLT recipients. This underlines the importance of an adequate post-transplant BMI on long-term survival after OLT.
OLT recipients have approximately 20% reduced survival rates when compared to the general population [30]. In the current study, this survival rate was comparable for normal weight OLT recipients. Obese OLT recipients, however, had an additional 20% decrease in survival rate. Moreover, obese OLT recipients have a 2-fold higher risk of mortality compared to normal weight OLT recipients, which is substantially more than obese people in the general population when compared to normal weight people (HR = 1.18, 95% CI: 1.12-1.25) [31]. This emphasizes the importance of a healthy weight for OLT recipients.
In the general population, a high BMI is associated with the development of cardiovascular diseases, diabetes, musculoskeletal disorders and cancer, resulting in increased morbidity and mortality [16]. Obesity is the hallmark of metabolic syndrome, of which hypertension, hypertriglyceridemia, hyperglycaemia, and low serum high density lipoprotein (HDL) cholesterol are other components [32]. Our study reveals that   obese OLT recipients have indeed more often hypertension, higher serum triglycerides, and higher glucose levels, when compared to OLT recipients with normal weight at 1 year post-transplantation. Interestingly, we found no significant association of BMI with cardiovascular mortality. We hypothesize that cardiovascular mortality is at least partly prevented because OLT recipients are periodically screened for cardiovascular risk factors (e.g. blood-pressure, serum glucose and cholesterol) on follow-up and adequately treated if necessary, when compared to the general population [33]. In our study, BMI was different across categories of primary liver disease, indicating that primary liver disease is associated with BMI 1-year post-transplantation. However, as demonstrated in model 3, BMI remained significantly associated with mortality despite adjustment for primary liver disease, indicating that high post-transplant BMI is associated with long-term mortality, independent of primary liver disease.
In this study, a BMI ≥ 30 kg/m 2 was associated with a significantly higher mortality risk, but we did not stratify obese OLT recipients into further categories of obesity. Nevertheless, Fig. 3 depicts an exponential relation between BMI and all-cause mortality in OLT recipients. Thus, further research is warranted to investigate potential differences in mortality risk in OLT recipients with a BMI N 35 kg/m 2 and BMI N 40 kg/m 2 post-transplantation.
Previous studies demonstrated that time-dependent BMI or increase in BMI post-transplantation is associated with better patient survival, which is inconsistent with our study results [47,48]. This might be explained by differences in study design (e.g. change in BMI, timedependent BMI, or BMI 1 year post-transplantation was used), differences in multivariable analyses, or the exclusion of patients with a BMI N 25 kg/m 2 in one study [48].
No significant differences were found in urinary CER between BMI groups. Therefore, in our study, weight gain after OLT was mostly due to an increase in fat mass, whereas muscle mass was independent of body weight. This increase in fat mass can be accelerated by poor lifestyle factors, including dietary intake, reduced physical activity, and immunosuppressive medication [20,23,49]. Although there is evidence that physical exercise improves long-term quality of life after OLT [50], studies on nutritional and physical-activity based interventions on long-term survival after OLT are currently lacking. Therefore, future studies concerning the effects of dietary advice and physical exercise on long-term survival are warranted. Furthermore, in selected patients, bariatric surgery, in particular sleeve gastrectomy, might be feasible and results in weight loss [51,52]. More research, however, is warranted to minimize risk of complications after bariatric surgery in OLT recipients. Optimal timing of bariatric surgery for obese recipients (i.e., before, during, or after OLT), remains to be defined [51,52].
This study has some valuable strengths. This study is characterized by an excellent follow-up. Median follow-up was 12,3 years (IQR 8.4-15.0 years) and only 1 OLT recipient was lost to follow-up. Additionally, only 13 OLT recipients (3%) were excluded because of missing baseline data on BMI or urinary CER. Furthermore, potential confounders were thoroughly addressed by using appropriate statistical analyses.
The current study has some limitations. The external validity of its findings is limited, due to the single-center cohort design. Furthermore, the post-hoc character of this study relies on adequate weight and height measurements performed by health care professionals, although we do not expect substantial variability among measurements.
In conclusion, post-transplant BMI is inversely associated with longterm survival after OLT. Moreover, obesity at 1-year post-transplantation conveys a 2-fold higher mortality risk, which may offer potential for interventional strategies (i.e. dietary advice, lifestyle modification, or bariatric surgery) to improve long-term survival of obese OLT recipients.

Declaration of competing interest
The authors of this manuscript have no conflicts of interest to disclose.