Introduction
Chronic kidney disease (CKD) is a common and significant complication after orthotopic liver transplantation (OLT). With the progression of surgical technology, perioperative treatment, and immunosuppressants, the survival rates of patients after OLT have dramatically improved to 90 and 75% at 1 and 5 years after OLT, respectively [
1]. Long-term complications, such as CKD and cardiovascular events, are receiving increasing attention from clinical specialists. The prevalence of CKD reported in the literature was 4 ~ 27.5% in 1 year [
2‐
4], 15 ~ 60% in 5 years [
2,
4,
5], and 25 ~ 50% in 10 years after OLT [
6,
7]. CKD is a primary contributor to morbidity and mortality after OLT, including an increased risk of allograft dysfunction, cardiovascular events [
8], and death [
4,
9], which has elicited concern regarding post-OLT management. However, only a limited number of studies have assessed long-term renal outcomes after OLT, which has important implications for patient management.
Studies have shown that the causes of CKD after OLT are complex and may include preoperative (age, female sex, history of diabetes mellitus (DM) or hypertension, hyperlipidaemia, and hepatitis C) and postoperative (calcineurin inhibitor (CNI) toxicity, prolonged ischaemia, and haemodynamic instability) factors [
6,
10‐
13]. There have been controversial results regarding a direct association between perioperative AKI and post-OLT CKD [
14,
15].
A credible and helpful predictive model to identify which patients are at high risk of developing CKD in clinical practice has not been developed, although it would be a useful tool for designing preventative measures, such as early decreased doses or withdrawal of CNIs and administration of non-nephrotoxic immunosuppressants. In this study, we aimed to develop and validate a reliable nomogram model for predicting CKD after OLT. With this model, we can identify patients at high risk for CKD early and take targeted measures to prevent or slow the development and progression of CKD.
Discussion
The increase in the number of long-term survivors of OLT raises questions about the increasing number of patients who develop CKD and ESRD. In our study, 25.6% of 399 recipients developed CKD at 5 years after OLT, which was consistent with that in previous reports as described earlier in this article. The survival curve showed that post-OLT CKD was a strong predictor of patient prognosis, including mortality. The time-dependent changes in renal function indicated that eGFR declined slowly year by year. It is important to identify patients at risk for CKD, especially in the early course, to slow the occurrence and progression of CKD.
Although different studies have attempted to identify the best predictors of renal dysfunction after liver transplantation, a reliable and simple model to accurately predict CKD is still lacking. For this reason, we chose to construct and validate a monogram model based on Cox regression analysis. A nomogram model can integrate multiple factors, enabling clinicians to comprehensively predict the prognosis of a patient and make clinical decisions. The model developed in our study integrated patients’ age at surgery, female sex, preoperative hypertension, baseline eGFR, UA, Hb, and average plasma concentration of CsA at 3 months after OLT as predictors of the development of CKD.
The first predictor in our model was female sex, which has previously been recognized as a risk factor for CKD. Female patients suffered a faster decline in renal function, as illustrated by the plot of the decrease in eGFR between males and females in this study. In contrast, among nontransplant patients, male sex was associated with a faster decline in eGFR. Female sex is also a risk factor for CKD after heart, lung, and kidney transplantation. The exact mechanism of this association is not clear. A possible explanation may be that females may be more susceptible to CNI-mediated renal injury [
21].
Recipients with hypertension and low eGFR are more prone to suffering IRI during and after surgery due to trauma and circulation instability [
6]. Although their
p values were > 0.05, our nomogram model showed that preoperative hypertension and low eGFR were important predictors of new-onset CKD after OLT.
In our population study, the UA level at 3 months after OLT was found to be a risk factor for new-onset CKD. Elevated serum levels of UA are associated with the onset and progression of CKD in many populations, such as the general population [
22], patients with hypertension [
23], and kidney transplantation patients [
24]. Although hyperuricaemia is common after OLT, its association with the development of CKD has not been well described. It was reported that post-OLT hyperuricaemia was related to immunosuppressants, primarily CNIs (CsA and FK506), which decrease UA excretion by reducing GFR and increasing UA reabsorption in renal tubules. The incidence of hyperuricaemia was much higher in recipients treated with CsA than in those treated with FK506 in some reports. Multiple mechanisms are related to renal damage caused by UA, including contributions to the development of hypertension [
25], activation of the renin-angiotensin-aldosterone system (RAAS) [
26‐
28], inflammation, and oxidative stress reactions [
29]. In patients with CKD, a high level of serum UA was associated with cardiovascular diseases [
30], a higher risk of incident RRT, and all-cause mortality [
31]. It is important to reduce UA levels in OLT recipients to prevent the occurrence of CKD and improve prognosis. The treatments include changing the unhealthy lifestyle of patients and using drugs such as febuxostat, allopurinol, and topiroxostat. In addition, CNIs should be avoided, or the dose should be reduced as much as possible. Some reports have shown that the reduction or even withdrawal of CNIs after OLT combined with MMF or mTOR inhibitors can contribute to the reduction in UA levels [
32,
33]. Therefore, serum UA levels should be closely monitored by clinicians during follow-up. Immunosuppression programs should be individualized according to the specific situation of the recipient. Due to its negative impact on prognosis, serum levels of uric acid were monitored and intervened carefully for all patients in our centre. The HR of UA at 3 months was low in this study, which might be related to the intervention of lowering UA during follow-up.
Another risk factor for CKD after OLT was low Hb levels at 3 months. The association between preoperative Hb and CKD after OLT was found in a previous Chinese cohort study [
6]. In our study, low Hb at 3 months after OLT may contribute to CKD by reducing the oxygen capacity of the blood, enhancing oxidative stress, and impairing haemostasis. In addition, a recent study from Korea revealed that CKD patients with anaemia are at high risk for hyperuricaemia, which may further promote the development of CKD. The reason for the decrease in Hb at 3 months may be related to the use of immunosuppressive medications. MMF, azathioprine, and sirolimus are common causes of bone marrow suppression, which thus leads to a decrease in Hb after OLT [
34].
It is well known that the dose-dependent nephrotoxicity of CNIs also plays an important role in the occurrence of new-onset CKD. In our study, the average plasma concentration of CsA at 3 months was an independent risk factor for the development of post-OLT CKD. Calcineurin stimulates vascular endothelial cells to secrete endothelin, release angiotensin II and overexpress transforming growth factor-β. This process is accompanied by a decrease in stromal degrading enzyme activity, leading to the hyperconstruction of glomerular arterioles, hyalinosis, chronic thromboembolism, and the excessive synthesis of extracellular matrix. Finally, it results in tubule atrophy and interstitial fibrosis and reduces renal blood flow and glomerular filtration [
6]. However, the concentration of tacrolimus was not included in the nomogram model. Most of our recipients were on long-term immunosuppression with tacrolimus, and interestingly, CKD recipients had significantly lower levels of tacrolimus. This finding was consistent with the finding of some previous studies, which did not show a correlation between tacrolimus levels and the development of CKD [
35]. This may be due to the adjustment of tacrolimus dose in patients with known renal impairment in our centre.
Finally, a nomogram model was constructed based on multivariable Cox regression analysis, which provides clinicians with a visual tool to understand the impact of predictors on renal function after OLT. In addition, by comprehensively analysing all predictors included, we can accurately calculate the probabilities of CKD for each recipient, making the results more personalized. After evaluation, the C-index of the nomogram model was 0.75 for the training set and 0.80 for the validation set, indicating that the nomogram model has a strong ability to distinguish patients with and without CKD. The calibration curves in the training set and validation set showed that the predicted probabilities of CKD at 1, 3, and 5 years after OLT were consistent with the observed probabilities, which indicates the accurate prediction ability of our model.
Several prediction models for CKD after OLT have been developed. A CKD prediction formula was developed from New York based on urinary neutrophil gelatinase-associated lipocalin (uNGAL) at 24 h after OLT as the most important risk factor. However, uNGAL-24 h is not routinely examined in other centres. Moreover, this prediction model has not been validated in terms of its calibration abilities [
36]. Recently, a biomarker model predictive of renal outcomes after liver transplantation was constructed based on a large population from multiple centres. The levels of β2-microglobulin and CD40 antigen included in the prediction model are not routinely examined in other centres. Despite the high area under the curve (AUC), no calibration verification was performed [
37]. The pocket guide to identifying patients at risk for CKD after liver transplantation, which includes hepatitis C virus (HCV) as one of the predictive factors, is a reliable prediction model for German patients. However, it does not apply to Chinese patients because HCV is more common in Western countries than in China [
38]. Another model from Italy has been validated through AUC calculation and calibration; however, the sample size was relatively small, and there was no visual tool developed for the model [
2]. In addition, none of these prediction models have been used in Chinese recipients.
Compared with these models, our model has the following advantages. First, it is the first nomogram model constructed based on OLT recipients for predicting CKD. Second, the predictors included in our model were readily available at surgery or routinely tested in follow-up, which is conducive to the popularization of the model. Third, our model had excellent discrimination and calibration abilities in the training set and validation set. It is a simple and reliable tool to distinguish recipients at high risk for CKD after OLT at 1, 3, and 5 years. In addition, this model can be used earlier at 3 months after OLT, assisting clinicians in adjusting immunosuppressive drugs in advance.
This study had several limitations. First, this study had a retrospective single-centre design, and the sample size was small, so the model should be evaluated prospectively in other large multi-centre studies to demonstrate its applicability. Second, the use of MDRD to assess renal function may lack accuracy in candidates awaiting liver transplantation, possibly because of reduced muscle mass and/or hyperbilirubinemia. It would overestimate renal function in patients with poor function (GFR < 40 mL/min) and underestimate renal function in patients with reasonable renal function (GFR > 40 mL/min) [
17]. However, this is the most frequently used method to estimate kidney function. Third, due to data deficiency, we did not analyse the relationship between urinalysis data such as proteinuria and postoperative CKD. Fourth, to more accurately clarify the impact of CKD on the long-term prognosis after OLT, it is necessary to conduct a longer follow-up study.
In conclusion, the renal function of most recipients decreased and recovered rapidly in the first week after OLT and then decreased slowly year by year. Patients with severe CKD had a poor survival prognosis. We constructed a nomogram model for predicting post-OLT CKD for the first time. With excellent discrimination and calibration, this nomogram model can accurately predict patients at high risk for CKD after OLT. Therefore, we can take measures to prevent or slow the development and progression of CKD in advance, such as reducing UA levels, improving Hb levels, and withdrawing or minimizing the use of CNIs in follow-up.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.