Hepatic Fibrosis Evaluated in Patients with Type 2 Diabetes with and without Chronic Kidney Disease

Hepatic fibrosis has been shown to be the most predictive factor for severe liver complications and mortality [1, 2]. Fibrosis is part of the non-alcoholic fatty liver disease (NAFLD) spectrum which ranges from simple steatosis to non-alcoholic steatohepatitis (NASH) with or without fibrosis which can progress to overt cirrhosis [3, 4]. NAFLD, especially in its severe forms, has been suggested as an independent risk factor for chronic kidney disease (CKD) [5]. Also, fibrosis seems to be an independent risk factor for reduced kidney function in people with preserved kidney function [6‒9] as well as in those with estimated glomerular filtration rate (eGFR) >30 mL/min/1.73 m² [10]. However, the association between liver fibrosis and CKD stages 3–5 is poorly elucidated.

It is well known that diabetes represents an important risk factor for NAFLD. It is presumably a bidirectional relationship with insulin resistance as a common factor [11]. Persons with type 2 diabetes have a 2-fold increased risk of having NAFLD compared with people without type 2 diabetes [12]. Further, persons with type 2 diabetes seem to progress faster and are therefore at higher risk of developing NASH [13, 14]; hence, the prevalence of NASH in patients with type 2 diabetes is 37% compared with 17% in the general population [15]. Complications of diabetes also include diabetic kidney disease, occurring in around 20% of patients with type 2 diabetes [16, 17], making diabetes the leading cause of CKD [18].

Recently, two studies from our group showed similar prevalence of hepatic steatosis in patients with CKD as in those without CKD [19, 20]. Using computed tomography to diagnose NAFLD, similar prevalence of moderate-to-severe hepatic steatosis was found in 291 patients with CKD stages 1–5 as in 866 age- and sex-matched individuals from the Danish background population [19]. Likewise, using proton magnetic resonance (MR) spectroscopy (¹H-MRS) and MR imaging (MRI) proton density fat fraction (MRI-PDFF) as diagnostic tools for NAFLD, we found a similar prevalence of hepatic steatosis in patients with type 2 diabetes with and without CKD (N = 100) [20].

Here, we investigated the association between liver fibrosis and kidney function in those patients participating in our latter study (N = 100), in which a total of 41% were diagnosed with NAFLD [20]. Liver fibrosis was assessed by transient elastography (FibroScan^®), fibrosis-4 (FIB-4) score, and NAFLD fibrosis score (NFS), respectively. As NAFLD and CKD seem to share common proinflammatory and profibrotic mechanisms of disease progression [21], we hypothesised that liver fibrosis was more prevalent in patients with CKD compared with those without.

Details on study participants, materials, and methods have previously been described [20]. In brief, this was a cross-sectional study carried out at the Department of Nephrology at Copenhagen University Hospital – Rigshospitalet, and at the Danish Research Centre of Magnetic Resonance (DRCMR), Copenhagen University Hospital – Amager and Hvidovre, Hvidovre, Denmark, between June 2019 and June 2021. Inclusion criteria for both groups were age 18–90 years and a diagnosis of type 2 diabetes. Exclusion criteria were use of steatogenic drugs; known viral, inherited, or alcoholic liver disease; or daily alcohol consumption ≥30 g and ≥20 g for men and women, respectively. Further, standard MRI contraindications for the strength of 3 Tesla as well as an upper body mass index (BMI) limit of 35 kg/m² and/or a measurement of the supine abdominal height ≤30 cm were used as exclusion criteria due to the size of the MRI scanner. A total of 100 patients were included: 50 patients with type 2 diabetes and CKD stages 3–5 (no dialysis) and 50 patients with type 2 diabetes and normal kidney function without clinical signs of diabetic nephropathy. NAFLD was assessed by ¹H-MRS and MRI-PDFF, as described in detail elsewhere [20].

Information about medication, medical history, and alcohol intake was obtained through patient interviews and electronic medical records. We recorded the following clinical parameters: blood pressure; height and weight; and hip and waist circumference. Urine and blood samples were collected after a minimum of 10 h fasting.

Hepatic fibrosis was assessed by three methods: FibroScan (502 Touch; Echosens, Paris, France), FIB-4 score, and NFS. Liver stiffness was measured by FibroScan by two examiners {bias 1.0% (limits of agreement [LoA] −35.8%; 56.3%) and −4.3% (LoA −35.8%; 56.3%), respectively, p = 0.3402}. This was performed with the patient lying supine with the right arm elevated to facilitate access to the right liver lobe [22]. A successful examination was defined as: (1) ≥10 valid measurements; (2) a success rate (the ratio of valid measurements to the total number of measurements) > 60%; and (3) an interquartile range (IQR) < 30% of the median value [22]. We aimed to make duplicate determination in all patients. The M-probe was used in all patients but two (both with CKD), in whom the XL-probe was used instead. The results are expressed in kilopascal (kPa) and correspond to the mean median value of the two valid measurements. The patients were divided into the following two categories: <8 kPa: absence of fibrosis or mild fibrosis (F0-F1); and ≥8 kPa: significant fibrosis (F2-F4) according to current guidelines [23, 24]. The FIB-4 score was calculated using the formula: (age [years] × aspartate aminotransferase [AST] [U/L])/(platelets [10⁹/L] × √alanine aminotransferase [ALT] [U/L]) [25]. Three risk categories for FIB-4 were calculated: ≤1.3 (low); between 1.3-2.67 (intermediate); and ≥2.67 (high) based on the 2 cut-off points described in the original publication [26]. The following equation was used to calculate the NFS: (−1.675 + 0.037 × age [years] + 0.094 × BMI [kg/m²] + 1.13 × IFG [impaired fasting glucose] or diabetes [yes = 1, no = 0] + 0.99 × AST/ALT ratio – 0.013 × platelet count [×10⁹/L] – 0.66 × serum albumin [g/dL]) [27]. Likewise, three risk categories were calculated: ≤−1.455 (low); between −1.455–0.676 (intermediate); and ≥0.676 (high) in accordance with the original study [27]. Thus, using single-test analyses, significant fibrosis was defined as either ≥8 kPa assessed by FibroScan^®, ≥2.67 assessed by the FIB-4 score, or ≥0.676 assessed by the NFS [24, 26, 27]. Patients who had an increased liver stiffness measurement ≥7 kPa and/or increased FIB-4 score ≥1.45 (patients <65 years) or ≥2 (patients ≥65 years [28]) were invited to be referred to the Gastro Unit, Medical Division, Copenhagen University Hospital – Amager and Hvidovre Hospital, Hvidovre, Denmark for further assessments by a hepatologist.

The CKD stages were classified according to KDIGO [29] and using the CKD-EPI formula [30]. Normal kidney function was defined as eGFR ≥60 mL/min/1.73 m² and urine albumin-to-creatinine ratio ≤300 mg/g. The definition of type 2 diabetes agreed with the World Health Organization (WHO) [31], and the definition of metabolic syndrome followed the recommendations made by the International Diabetes Federation (IDF) [32].

Continuous data with a symmetric distribution were presented as mean ± SD, whereas skewed continuous data were presented as median (IQR). Categorical data were presented as n (%). Variables showing log-normal distribution were analysed on logarithmic scale. Unpaired t test or Mann-Whitney U test was used for comparison of continuous data; and χ² or Fisher’s exact test was used for comparisons of categorical data. Differences were quantified using the difference of means, the difference of proportions, or the Hodges-Lehmann estimator of difference in location, respectively. All these estimates of group difference were supplemented by the corresponding 95% confidence intervals. Beyond single-test analyses, we performed paired combinations of liver stiffness measurements, FIB-4 score, and NFS for the diagnosis of significant fibrosis as recommended by Petta et al. [33] to improve accurate diagnosis of fibrosis. Only patients with complete biochemical data and liver stiffness measurements were included in the latter analyses. Bias and LoA for the measurements by FibroScan were analysed using Bland-Altman method. All statistical analyses were performed using SAS Enterprise Guide version 7.15 (SAS Institute, Cary, NC, USA). A p value <0.05 was considered statistically significant.

The mean age was 72 ± 4.9 years and 65.9 ± 7.8 years in patients with and without CKD, respectively (p < 0.0001). Three quarters of the study population were men. Patients with CKD had a mean BMI of 28.6 ± 3.5 kg/m², and patients without CKD had a mean BMI of 27 ± 4.0 kg/m² (p = 0.0087). Duration of type 2 diabetes, levels of HbA1_c, and presence of metabolic syndrome were similar in the two groups. NAFLD was defined in 22 (44%) and 19 (28%) of the patients with and without CKD, respectively. Further characteristics of the study population are listed in Table 1.

Liver stiffness measurements using FibroScan were successful in 47 patients with CKD and in all 50 patients without CKD. Significant fibrosis (≥F2) was identified in 5 of 47 (10.6%) patients with CKD and in 8 of 50 (16%) of the patients without CKD (p = 0.44) using FibroScan. No association was found between liver stiffness (kPa) measured by FibroScan and kidney function (eGFR) using a regression model. Unadjusted FIB-4 score showed significantly higher median values among patients with CKD compared with patients without CKD as shown in Figure 1 (1.59 (IQR: 1.27–1.9) versus 1.21 (IQR: 0.94–1.60), p <0.0009). The difference became insignificant after adjustments for age, sex, BMI, and duration of diabetes (p = 0.1929). Considering eGFR as a continuous variable resulted in a significant linear association between eGFR and FIB-4 score, with a 10% decrease in eGFR increasing the FIB-4 score with 1.8% (p = 0.0130), adjusted for age, sex, BMI, and duration of diabetes. Unadjusted median (IQR) NFS were 0.21 (−0.39; 0.96) and −0.53 (−1.49; −0.12) in patients with and without CKD, respectively (p < 0.0001) as shown in Figure 1. After adjustments for age, sex, BMI, and duration of diabetes, the difference was insignificant (p = 0.1680). When considering eGFR as a continuous variable, a decrease in eGFR of 10 mL/min/1.73 m² increased the NFS with 0.08 score units when using a linear regression model adjusted for age, sex, BMI, and duration of diabetes (p = 0.038). In comparison, using the same linear regression model, NFS increases by 0.06 score units per increased year of age (p < 0.0001) and by 0.10 score units per increased BMI unit (p < 0.0001). In total, significant fibrosis measured by FibroScan, FIB-4 score, or NFS was identified in 19 (38%) of the patients with CKD and 14 (28%) of the patients without CKD (p = 0.3952). When combining measurements obtained by FibroScan with FIB-4 score or NFS, 2.0% and 6.1% of the total study population, respectively, had significant fibrosis as specified in Figures 2 and 3. Agreement between all three methods was only observed in patients without CKD as shown in Figure 4. In total, 30 patients were referred to a hepatologist. Of these, 4 patients had a percutaneous liver biopsy performed, in whom NASH with different stages of fibrosis was diagnosed in 3 patients (online supplementary material Table S1; Fig. S1; for all online suppl. material, see https://doi.org/10.1159/000531574).

In patients with CKD, those with significant fibrosis had increased weight and BMI compared with patients without significant fibrosis. In patients without CKD with significant fibrosis, all liver parameters were increased, and all fulfilled the criteria for metabolic syndrome including increased weight and BMI. Cardiovascular events were more frequently occurring in patients with significant fibrosis as in those without. Further characteristics are presented in Table 2.

In patients with CKD, 7 of 22 (31.8%) of those with NAFLD defined by ¹H-MRS or MRI-PDFF and 12 of 28 (42.9%) of those without NAFLD fulfilled the criteria of significant fibrosis using one of the three methods (p = 0.5595). In patients without CKD, 11 of 19 (57.9%) and 3 of 31 (9.7%) of patients with and without NAFLD defined by ¹H-MRS or MRI-PDFF, respectively, fulfilled the criteria of significant fibrosis (p = 0.0007), and statistical significance was observed using any of the three methods: FibroScan (p = 0.0427), FIB-4 score (p = 0.0005), and NFS (p = 0.0032). Considering steatosis as a continuous variable, significant associations between liver fat fraction (%) and FIB-4 score (p = 0.0044), and NFS (p = 0.0054) were observed in patients without CKD when data were adjusted for age, sex, and BMI. Corresponding analyses were not significant in patients with CKD (P_{FIB-4 score} = 0.5862, P_NFS = 0.9975). Adjusted data showed significant associations between liver fat fraction (%) and liver stiffness (kPa) measured by FibroScan both in patients with CKD (p = 0.0133) and in those without CKD (p = 0.0031).

In the present study, we found a similar prevalence of liver fibrosis in patients with type 2 diabetes with and without CKD. Although both FIB-4 score and NFS were significantly higher in patients with CKD, the differences were insignificant after adjustments for age, sex, BMI, and duration of diabetes.

Significant fibrosis was identified in a considerable proportion of this study population without known liver disease prior to inclusion. When screening for liver fibrosis, the choice of method for assessment is essential. The demonstrated prevalence of significant fibrosis or that of having a high risk of significant fibrosis by at least one of the three methods of 38% and 28% in patients with and without CKD, respectively, agrees well with reported estimations of the prevalence of NASH in patients with type 2 diabetes [12]. Using FibroScan, the most widely used imaging technique worldwide for diagnosing hepatic fibrosis [23, 24], significant fibrosis was detected in 10.6% of the patients with CKD and in 16% of the patients without CKD, which is congruent with previously findings, among others by Mantovani et al. [34] and Roulot et al. [35]. In the first study, the prevalence of significant fibrosis in 137 patients with type 2 diabetes examined by FibroScan was 17.5% using cut-off ≥7 kPa and 10.2% using cut-off ≥8.7 kPa [34]; in the latter, 12.7% of a community-based French population with type 2 diabetes (N = 705) exhibited significant fibrosis defined as liver stiffness measurement ≥8 kPa. Finally, more accurate definitions of significant fibrosis using paired combination of FibroScan and either FIB-4 score or NFS resulted in a much lower prevalence: 2.0% and 6.1%, respectively. Moreover, 3 patients, corresponding to 3% of the total study population, were histologically diagnosed with NASH with different stages of fibrosis (ranging from F1 to F4). These data are reasonably equal to a large meta-analysis in which the prevalence of advanced fibrosis assessed by liver biopsy was found to be 4.8% in the general type 2 diabetes population [12]. Similar results were demonstrated by Doychera et al. [36] who identified advanced fibrosis in 7% of patients with type 2 diabetes using MR elastography. The three assessment tools, FibroScan, FIB-4 score, and NFS, are obviously all very good at ruling out and ruling in significant fibrosis [24, 26, 27]. Although all three methods have a high negative predictive value (>90%), the accuracy is a concern [37].

Liver stiffness measured by FibroScan associated significantly with increasing liver fat fraction estimated by ¹H-MRS or MRI-PDFF in the whole study population. Interestingly though, associations between measurements obtained by FibroScan and FIB-4 score as well as between steatosis and fibrosis were found in patients without CKD only. This could possibly be due to the association between CKD and NAFLD being complex, and that the co-existence of both diseases interferes in a way that is not fully understood by now. FIB-4 score and NFS are shown to be the best biomarkers indicating significant fibrosis [38, 39] and growing evidence indicates an association between these two and reduced kidney function in patients with NAFLD and preserved kidney function as baseline [6‒9, 40, 41]. In accordance, we observed an inverse weak association between continuous eGFR and FIB-4 score, and NFS, respectively. Two other studies have investigated fibrosis in patients with established CKD, though not on renal replacement therapy [10, 42]. In a study by Mikolasevic et al. [42], 62 patients with CKD were investigated by transient elastography, and no association between liver stiffness severity and eGFR was observed. On the contrary, Aubert et al. [10] used NFS to show that patients with type 2 diabetes and eGFR >30 mL/min¹/1.73 m² with high risk of fibrosis had a significant elevated risk of CKD progression compared with patients with low risk of developing hepatic fibrosis during 75.8 ± 23.9 months of follow-up including 102 patients. Hence, according to this study hepatic fibrosis seems to play a role in the development of CKD. Still, these results need to be confirmed. As both hepatic fibrosis and CKD progress into more severe stages over time, studies including the dynamic changes, for both diseases, are needed – regardless they are retro- or prospective.

To come a step further toward understanding a possible cross-link between NAFLD and the kidneys, focus could preferably be on patients with CKD stages 3–5 with severe stages of the NAFLD spectrum. To our knowledge, there are currently no studies involving patients with concomitant biopsy-proven NASH and biopsy-proven CKD which could allow for better understanding of the pathophysiological mechanisms linking liver and kidney disease. Possible explanations have, though, been proposed. These include, among others, insulin resistance, oxidative stress, and activation of nuclear factor kappa-light chain-enhancer of activated B cells (NF-kB) pathway that indirectly causes increased intra-hepatic cytokine production which might have a pathogenetic role in the development of CKD [43]. Further, ectopic lipid deposition contributing to local oxidative stress, inflammation, and fibrosis occur in both NAFLD and CKD [21, 44]. Nevertheless, any direct pathophysiological link between NAFLD and CKD has still not been found, as the proposed mechanisms, more or less, might be confounded by common risk factors [45, 46]. Hence, which implications NAFLD, and especially fibrosis, have in patients with CKD stage ≥3 remains to be clarified.

Our study has some limitations including a relatively small sample size and the cross-sectional design. Moreover, we did not perform liver biopsies in all patients which otherwise would have given us the opportunity to compare our results with histological data. Although liver biopsy remains the gold standard and the only reliable method by which NASH can be distinguished from NAFLD [47], it is an invasive procedure with potentially life-threatening complications; hence, taking a biopsy in a probably healthy individual involves great ethical considerations. We did not pursue further analysis of significant fibrosis using multiple logistic regression as we judged the number of cases with significant fibrosis to be too small. This could have been an interesting contribution to the results and a benefit of the design of future studies. The study also has some important strengths. To the best of our knowledge, this is the first study combining diagnosis of steatosis obtained by ¹H-MRS and MRI-PDFF with measurements of liver fibrosis obtained by FibroScan, FIB-4 score, and NFS in patients with type 2 diabetes and CKD stages ≥3. This allows us to compare and gain new insights into different methods for detecting significant fibrosis in this patient population.

To conclude, we found similar prevalence of fibrosis in patients with type 2 diabetes with or without CKD as assessed by FibroScan, FIB-4 score, and/or NFS. Overall, our data do not support any strong association between liver fibrosis and kidney function in patients with type 2 diabetes and established CKD.

We thank all participants for their contribution to the study. We thank the Department of Intestinal Failure and Liver Diseases at Rigshospitalet for allowing us to use their FibroScan. We also thank laboratory technician Andreas Haltorp and study nurse Helle Corinth from the Department of Nephrology for their valuable contributions.

The study followed the principles of the Helsinki Declaration of 1975, as revised in 2013, and informed written consent was obtained from all participants prior to any examination. The study protocol was approved by the Scientific-Ethical Committee of the Capital Region of Denmark (H-17040866) and The Data Protection Agency (P-2019-96). The study was registered at clinicaltrials.gov prior to recruitment (NCT03826381).

Therese Adrian, Mads Hornum, Karl Bang Christensen, Lisa í Lída, Vincent Oltman Boer, Anouk Marsman, Esben Thade Petersen, and Bo Feldt-Rasmussen have nothing to declare. Filip Krag Knop has served on scientific advisory panels and/or been part of speaker’s bureaus for, served as a consultant to, and/or received research support from the following companies producing and/or developing SGLT-2 inhibitors and/or GLP-1RAs mentioned in the manuscript: AstraZeneca, Boehringer Ingelheim, Eli Lilly, MSD/Merck, Novo Nordisk, and Sanofi. Thomas Almdal has received support for attending meetings from Novo Nordisk and Boehringer Ingelheim and holds stocks in Novo Nordisk. Peter Rossing has received consultancy and/or speaking fees (to Steno Diabetes Center Copenhagen) from Astellas, AstraZeneca, Bayer, Boehringer Ingelheim, Eli Lilly, Gilead, MSD, Mundipharma, Novo Nordisk, Vifor, and Sanofi Aventis; and research grants from AstraZeneca, Bayer, and Novo Nordisk. Niels Søndergaard Heinrich holds stocks in Novo Nordisk A/S and Akcea Therapeutics, Inc. Hartwig Roman Siebner has served on scientific advisory panels and/or been part of speaker’s bureaus for, served as a consultant to, and/or received research support from the following companies: Sanofi Genzyme, Lundbeck AS, Novartis, Lophora Aps, Elsevier Publishers, Danish Parkinson Association, and Danish Society for Medical Magnetic Resonance. He has received research grants from Independent Research Fund Denmark, The Novo Nordisk Foundation, and Lundbeck Foundation, and royalties as book editor from Springer Publishers, Stuttgart, Germany and from Gyldendal Publishers, Copenhagen, Denmark. The results presented in this paper have not been published previously in whole or in part, except in abstract format.

This study was funded by Novo Nordisk Foundation (Steno Collaborative Grant, NNF17OC0027944). The funding was not involved in the design of the study; the collection, analysis, and interpretation of data; or writing the report and did not impose any restrictions regarding the publication of the report.

Therese Adrian designed the study, participated in recruitment of the patients, performed examinations of the patients, performed and evaluated data analysis, wrote the first draft of the manuscript, critically revised and provided intellectual content to the work described, and approved the final version. Mads Hornum and Bo Feldt-Rasmussen designed the study, evaluated data, critically supervised the draft manuscript, provided intellectual content to the work described, and approved the final version. Filip Krag Knop contributed to the study conception and design of the study, critically revised the manuscript, provided intellectual content to the work described, and approved the final version. Karl Bang Christensen contributed with the statistical analyses and interpretation of data, provided intellectual content to the work described, and approved the final version of the manuscript. Thomas Almdal and Peter Rossing contributed to the study conception and design of the study, revised the manuscript, provided intellectual content to the work described, and approved the final version. Lisa í Lída participated in recruitment of the patients and performed examinations of the patients, revised the manuscript, provided intellectual content to the work described, and approved the final version. Niels Søndergaard Heinrich participated in recruitment of the patients, revised the manuscript, provided intellectual content to the work described, and approved the final version. Vincent Oltman Boer analysed and evaluated data from the MRS and MRI scans, revised the manuscript, provided intellectual content to the work described, and approved the final version. Anouk Marsman and Esben Thade Petersen contributed to the performance of the MRS and MRI scans and analysed and evaluated data from the MRS and MRI scans, revised the manuscript, provided intellectual content to the work described, and approved the final version. Hartwig Roman Siebner contributed to the performance of the MRS and MRI scans, revised the manuscript, provided intellectual content to the work described, and approved the final version.

The datasets generated during and analysed during the current study are not publicly available due to Danish legal restrictions but are available from the corresponding author on reasonable request, provided relevant ethical and legal permissions have been attained.