Changes in body mass index and outcomes after kidney transplantation: a single centre, retrospective, observational study

The aim of this study was to describe the changes in body mass index (BMI) after kidney transplantation and assess how this influences long-term outcomes. Methods Data were collected for all kidney transplant recipients between January 2007 and July 2016. Changes in BMI over the post-transplant period were modelled using a generalised estimating equation. The change in BMI from pre-transplantation to six months was then calculated for each patient. These were categorised into three groups: stable BMI (a change of ±1.5 kg/m2), BMI reduction and BMI increase (changes of >1.5 kg/m2), between which a range of outcomes were compared. Data was available for 1,344 patients, who had a geometric mean pre-transplant BMI of 27.3 kg/m2. This declined significantly (P<0.001), to a geometric mean of 25.6 kg/m2 one month after transplantation, before increasing and stabilising to pre-transplant levels by 36 months (geometric mean 27.2 kg/m2, P=0.522). The n=882 patients with BMI measurements at six months, were divided into groups of reduced (n=303), stable (n=388) and increased (n=131) BMI, relative to pretransplantation levels. On multivariate analysis, 12-month creatinine levels were significantly higher in the BMI reduction cohort, with adjusted levels of 160.6 μmol/l, compared to 135.0 μmol/l in stable BMI. However, no significant associations were detected between six-month BMI change and patient survival, graft survival, incidence of post-transplant diabetes, cancer, or a range of clinical and histological outcomes (all P>0.05). Our data demonstrates that BMI significantly reduces in the first month after kidney transplantation, before increasing to pre-transplant levels at 3-5 years. Furthermore, patients with decreasing BMI at six-months have impaired graft function in the long-term. These observations conflict with the existing literature and warrant further investigation.


Introduction
Post-transplant weight gain is a well-known occurrence in both obese and non-obese transplant recipients (1)(2)(3)(4). It has been documented in all forms of organ transplantation (5)(6)(7) and several studies have reported an average 5-10 kg increase in weight in the first year after kidney transplantation (2,4,(8)(9)(10). Weight gain may differ by geographic region, with an average increase in the first year after kidney transplantation of 2.7 kg documented in France versus 10.3 kg in America (2,11), and likely reflects the environmental confounders that affect weight gain among different cohorts. Post-transplant weight gain is significant, as it has been associated with inferior graft and patient survival (3) and is thought to contribute to cardio-metabolic risk profiles including hypertension (3), diabetes (3) and dyslipidaemia (12,13) and ultimately increased risk for cardiovascular-related death (14,15).
The aetiology of weight gain after kidney transplantation is multifactorial and likely influenced by individual, environmental and clinical factors. Immunosuppression, such as corticosteroids, has been shown to stimulate appetite, as well as induce glucose intolerance, hyperlipidaemia, hypertension, and impair vitamin D metabolism (2,4,16,17). However, immunosuppression with steroid-free protocols fails to significantly reduce this risk, with patients still experiencing weight gain (10,18). Additional influences on weight gain may relate to kidney transplant recipients with excellent graft function no longer being limited in their dietary restrictions that are advised to patients with advance kidney disease or failure (19,20).
While the impact of recipient or donor body mass index (BMI) on post-transplant outcomes has been detailed in the literature (21,22), the degree of 'risk' conferred by increasing or decreasing BMI after kidney transplantation among recipients is poorly understood. Furthermore, the prevailing literature that is present on this topic is outdated (pre-2001) and is published from centres in the United States.
Given the changes in transplant care that have occurred in the last decade, and well documented differences in clinical outcomes among kidney transplant recipients between Europe and the United States (23) (including degree of weight gain seen (11)), we believe a contemporary analysis in a European transplant cohort is warranted.
Therefore, the aim of this study was to explore the evolution of weight change after kidney transplantation and its link to clinical outcomes, to help transplant clinicians make evidence-based decisions in relation to weight amongst kidney transplant recipients.

Study population
Our analysis included all adult patients (aged 18 years of age and older) receiving a living or deceased donor kidney transplant between January 2007 and July 2016 (excluding recipients of multiple organs). We used BMI as our marker of change in weight, as it is the accepted measure for defining anthropometric height/weight characteristics in adults (24).

Outcome Measures
Our primary outcome measures were patient survival and death-censored graft survival. Secondary outcome measures included clinical outcomes comprising graft function (creatinine levels at 12-months' post-kidney transplantation), incidence of medical complications (e.g. post-transplant diabetes mellitus, cancer, cardiac events, cerebrovascular events, CMV viremia, urology complications, septicaemia and transplant artery stenosis) and histopathological findings. Since the change in BMI was measured over the six-month period post-transplant, follow up for these outcomes commenced at this time. Any patients where an outcome occurred prior to six months were excluded from the analysis of that outcome.

Immunosuppression protocol
All patients received the same immunosuppression as discussed in the SYMPHONY protocol, while reducing the amount of tacrolimus received (25). Patients received induction immunosuppression with Basiliximab and methylprednisolone.
To identify any dysfunction in the transplanted graft, biopsies were taken and scored according to the BANFF criteria (10). Treatment for patients included corticosteroids for acute rejection, with T-cell depletion if there was steroid resistant rejection. Patients with antibody-mediated rejection were treated with plasmapheresis +/-IVIg.

Statistical Analysis
We first modelled the changing BMI in the post-transplant period using a generalised estimating equation model. The timing of the measurement was set as the independent variable, and an AR(1) correlation structure was used to account for the correlations between repeated measures of BMI on the same patient. Since the BMI measurements followed a skewed distribution, the values were log 10transformed, prior to analysis, in order to normalise the distribution and improve model fit. The coefficients from the resulting model were then anti-logged, and converted into estimated geometric means with 95% confidence intervals (95% CI).
To assess the effect of weight change on post-transplant outcomes, we first A range of outcome measures were then compared across the BMI change groups, with Cox regression models for time to event outcomes and binary logistic regression models for dichotomous outcomes. Creatinine levels were found to follow a skewed distribution, and so values were log 10 -transformed in order to normalise the distribution, before being analysed using general linear models, whilst the numbers of biopsies were compared using a Kruskal-Wallis test. Multivariable analyses were then performed for each outcome, to account for the effect of confounding factors. Prior to this analysis, continuous factors were divided into categories, based on the tertiles of the distribution, in order to improve model fit.
All factors were then considered for inclusion in the model, with a backwards stepwise approach used to select those that were independently predictive of patient outcome. Where significant differences between the BMI groups were detected on multivariable analysis, the adjusted outcomes for each group were calculated. This was achieved by multiplying each coefficient by the proportion of patients the associated category, and evaluating the resulting model for each of the BMI groups. This gave the expected outcome for the "average" patient in the cohort, hence removing the impact of confounding factors.
All analyses were performed using IBM SPSS 22 (IBM Corp. Armonk, NY), with P<0.05 deemed to be indicative of statistical significance throughout.

Approvals
This study received institutional approval and was registered as an audit (audit identifier; CARMS-12578). The corresponding author had full access to all data.

Results
Trends in BMI During the study period, data were available for a total of N=1,387 transplant recipients, of whom N=1,344 (97%) had pre-transplantation BMI recorded. These patients had a geometric mean pre-transplantation BMI of 27.3 kg/m2 (95% CI: 27.0 -27.5), which was found to fall significantly by one-month post-transplantation, to 25.6 kg/m2 (95% CI: 25.3 -25.9, P<0.001). The average BMI was then found to increase progressively over the subsequent months ( Figure 1, Recipients with BMI reductions were conversely older and had a higher baseline BMI. They received organs from older donors and were more likely to receive kidneys from deceased donors (all P<0.001). Both waiting time and cold ischaemia time followed a 'U' shaped association, being highest in the BMI reduction and increase groups, and lower in the stable BMI group (P<0.001).
BMI change and outcomes On univariable analysis (Table 3) However, a significant difference in 12-month creatinine was noted (P<0.001), with median levels being higher in the BMI reduction group (144.0 μmol/l), and lower in the BMI increase group (117.5 μmol/l).
On account of the previously identified baseline differences between the three groups in recipient, donor and transplant related factors, multivariable analyses were performed in order to account for potentially confounding factors. The effect of the BMI change remained non-significant for all outcomes, except for renal function (Table 4), with the previously noted difference in 12-month creatinine levels remaining significant on multivariable analysis (P<0.001). After adjustment for confounding factors, creatinine levels at 12-months post-transplantation were found to be 19% higher in patients that had a reduction in BMI, compared to those with stable BMI. Evaluating the model at the midpoint of the confounding factors gave estimated adjusted creatinine levels of 160.6 μmol/l in the BMI reduction group to 135.0 μmol/l in the stable BMI group, and 131.0 μmol/l in BMI increase group.
We further conducted a subsidiary analysis of clinical and histological outcomes, the results of which are presented in Table 5. Overall, in univariable analysis, changes in six-month BMI were not associated with any risk of cardiac or cerebrovascular events, CMV viremia, septicaemia or transplant renal artery stenosis (all P>0.05).
Furthermore, no association between six-month BMI change and histological rejection was seen (all P>0.05).

Discussion
To our knowledge, this is the largest study to analyse the relationship between BMI  Results are from a generalised estimating equation model, with the timing of the measurement as the independent variable, and an AR(1) correlation structure used to account for correlations between repeated measurements on the same patient. The log 10 -transformed BMI measurements were set as the dependent variable, and the resulting coefficients were anti-logged, and converted into estimated geometric mean BMIs. P-values are in comparison to the pre-transplantation BMI, and Bold Pvalues are significant at < 0.05. Tx; Kidney Transplantation, CI; Confidence Intervals  All analyses were performed using a time-to-event approach using Cox regression models, with results summarised using hazard ratios and 95% confidence intervals (relative to the Stable BMI group), unless stated otherwise. Bold P-values are significant at P<0.05. DCGL; Death Censored Graft Loss. Tx; Transplantation. *Creatinine was found to follow a skewed distribution, and so was log 10transformed, and analysed using a general linear model, with the fold-differences between groups reported. **Patients with pre-Tx diabetes were excluded from the analysis of post-Tx diabetes.