Age and Serum Creatinine Can Differentiate Wilson Disease Patients with Pseudonormal Ceruloplasmin

Methods We retrospectively screened individuals with serum Cp ≥ 140 mg/L from 1032 WD patients who were hospitalised for the first time. Logistic regression analyses were performed in a case-control study between the WD cohort and another liver disease cohort to explore the independent risk factors for WD diagnosis and establish a regression model to identify them. The follow-up medical records of the WD cohort were subjected to mixed-effects model analysis in a longitudinal study to discover factors associated with Cp normalisation. Results Eighty-six WD patients and their 353 medical records and another 98 non-WD liver disease patients were included in the present study. Cp normalisation was significantly associated with the copper burden and liver function indexes, such as urinary copper, γ-glutamyltransferase, and albumin (p ≤ 0.001). Logistic regression analysis showed that age and serum creatinine (p ≤ 0.001) were independent risk factors associated with WD. The AUC value of the regression model in the total cohort was 0.926 (p ≤ 0.001). At a cutoff value of ≥0.617 and ≥−1, the positive and negative predictive values were both 90.8% for WD. Conclusion Increased serum Cp in WD patients is related to excessive copper burden and hepatic injury, and common tests can effectively distinguish WD patients from other liver injury patients.


Introduction
Wilson disease (WD) is an autosomal recessive copper metabolism disease caused by mutations in the ATP7B gene, which encodes a copper-transporting P-type ATPase to convey copper for synthesising ceruloplasmin (Cp) [1,2]. Te decrease in serum Cp concentration is a useful diagnostic hallmark of WD [2]. An epidemiological investigation revealed that the incidence and prevalence of WD in China were approximately 1.96/100,000 and 5.87/ kidney failure and end-stage liver disease will be just the opposite. Te normal concentration of Cp measured by the enzymatic assay has a lower limit of 200 mg/L [6]. Te diagnostic accuracy for WD using a cutof value of 140 mg/L had 100% positive predictive values and 97.1% negative predictive values [11]. In WD, it is usually lower than 100 mg/L, especially in neurologic WD. However, it may be in the low normal range in hepatic WD with active liver disease [12][13][14] and correlated with liver histologic activity [15]. Because these WD patients' manifestations and serum Cp overlap with those of other liver diseases, their diagnosis may be missed.
Some laboratory fndings and indexes are helpful in distinguishing WD with acute liver failure (ALF) from non-WD ALF, including increased aspartate aminotransferase: alanine aminotransferase (AST : ALT) ratios, low serum alkaline phosphatase (AP) activity, and decreased AP to total bilirubin (TB) ratios [13,16,17]. However, in the non-ALF WD patients with nearly normal Cp, the indicators that are diferent from those of other etiologies of hepatopathy have not yet been studied.
Here, we retrospectively screened 1032 WD patients, enrolled 86 individuals who had nearly normal serum Cp, and analysed their subsequent multiple medical records to identify factors related to serum Cp pseudonormalisation. In addition, according to logistic regression analysis, we compared the control group to develop a regression model to diferentiate WD from other hepatopathy patients [18].

WD Patient Cohort.
We retrospectively investigated 1032 WD patients who were frst hospitalised in the afliated hospital of the Neurology Institute of the Anhui University of Chinese Medicine (Hefei city, Eastern China) from March 2010 to July 2013. Te enrolled criteria included (1) meeting the Leipzig criteria for WD diagnosis [16,19], (2) serum Cp concentration being ≥140 mg/L at diagnosis [11], and (3) patients being younger than 65 years. Te following cases were excluded: (1) pregnancy, liver [20] or renal failure [21], chronic kidney disease, infection, cancer, and alcohol abuse; (2) other suspicious hepatitis such as viral or autoimmune hepatitis and drug-induced liver injury; and (3) incomplete medical records.
Finally, 86 WD patients (group A) were enrolled in this study. We followed up all of their electronic medical information as of December 2019 and obtained 352 records. To understand which indicators are associated with changes in serum Cp, in light of the serum Cp concentration, 352 medical records were classifed as group A1 (Cp ≥ 140 mg/L, n = 231) and group A2 (Cp < 140 mg/L, n = 121) to carry out a longitudinal study, regardless of which patient they belonged to. Te fow chart is shown in Figure 1.
Teir usual information and laboratory results, including sex, age, duration of onset, and follow-up, phenotype, Kayser-Fleischer ring (K-F ring), laboratory biochemical fndings, serum Cu and Cp, serum ceruloplasmin oxidase (Sco), 24-hour urinary copper, Child-Pugh scores and classifcation, abdominal ultrasonographic or radiological fndings, liver cirrhosis or not, and mutational sites of the ATP7B gene, were recorded. ATP7B was detected with nextgeneration DNA sequencing. AfterQC [22] was applied to generate "clean reads" for further analysis. Ten, the "clean reads" (with a length of 150 bp) were aligned to the human genome reference (hg19) using Burrows Wheeler Aligner (BWA) software [23]. After alignment, SNVs and indels were detected using Genome Analysis Toolkit (GATK) software [24]. All SNVs and indels were fltered and estimated via multiple databases, including 1000 Genomes (1000 human genome dataset), gnomAD (Genome Aggregation Database) dataset, and ExAC (Exome Aggregation Consortium) dataset. Serum copper was determined by using fame atomic absorption spectrophotometry [25]. Serum Cp was measured by the immunonephelometric assay with antiserum to human Cp [26]. Sco activity with odianisidine dihydrochloride as a substrate was determined by spectrophotometry [26,27]. Te basic 24-hour urinary copper and with intravenous infusion of dimercaptopropane sulfonate (DMPS) sodium or oral penicillamine was recorded.

Case-Control Cohort of Non-WD Liver Disease Patients.
WD patients with Cp normalisation would overlap with that of other liver patients, so we included the control cohort of non-WD liver disease patients to conduct logistic regression analysis and establish a model to distinguish WD patients.
Te control group patients came from the liver disease department of First Afliated Hospital, Zhejiang University School of Medicine, in March 2021. Tose patients were included if they met the following criteria: (1) identifable acute liver diseases and compensated or decompensated liver cirrhosis (LC) due to varied non-WD causes, (2) serum Cp concentration ≥140 mg/L, and (3) <65 years. Te exclusion criteria included pregnancy, liver or renal failure, chronic kidney disease, infection, and cancer. Finally, 98 liver injury patients were included in this control cohort, including 23 nonalcoholic steatosis hepatitis (NASH), 27 acute hepatitis (hepatitis B: 11, hepatitis E: 7, medicamentous: 5, autoimmune hepatitis: 2, and Epstein-Barr virus infection and hyperthyreosis: 1 for each), 4 compensated LC induced by hepatitis virus B (HBV), and 44 decompensated LC (induced by CHB: 38, alcohol: 2, and CHB + alcohol, medicamentous, primary sclerosing cholangitis, and Budd-Chiari syndrome: 1 for each).

Statistical Analysis.
All statistical analyses were performed with SPSS version 21.0 (IBM, New York, USA). Te chi-square and nonparametric tests, t-test, and ANOVA were carried out, respectively, depending on diferent data types. A generalized linear mixed-efects model (GLMM) was conducted on repeated measurement data (group A1 vs. A2, Figure 1) to identify indicators related to an increase in Cp. WD and non-WD case-control cohorts (group A vs. B, Figure 1) were divided into training and validation sets for conducting logistic regression analysis to develop a model and validate it. Factors that revealed signifcant diferences in the univariate analysis (p < 0.05) were then included in the multiple binary logistic regression analysis with the backward method to identify independent factors associated with WD and develop a regression model. Te area under the receiver operating characteristic curve (AUC) was used to evaluate the diagnostic value of the logistic regression model. Te sensitivity, specifcity, positive predictive value (PPV), and negative predictive value (NPV) of the model were calculated to determine the optimal and handy cutof values.

Baseline Characteristics of 86 WD Patients.
Te baseline characteristics and laboratory fndings of 86 enrolled WD patients (male: 42, 48.8%) at the frst admission are shown in Table 1. Tere were three phenotypes: hepatic type (H-type, n � 51, 59.3%), hepatic and neurologic presentations (HNtype, n � 27, 31.4%), and neurologic type (N-type, n � 8, 9.3%). None of the patients converted to other types during the follow-up period. Comparing the three phenotypes, there were no diferences in age at presentation (p � 0.091), but the age at diagnosis especially varied between them (p � 0.005). Terefore, the N-type had the longest undiagnosed duration and duration of onset, 12.5 ± 10.6 (mean ± standard deviation, the same below) years (p � 0.010), which is the time from onset to diagnosis. In addition, the duration and times of follow-up were 4.9 (1.8, 7.1) (median, quantile, the same below) years and 4 (2, 6) times, varying from 0.5-9.8 years and 1-10 times, respectively. 28 of their serum Cp concentrations were ≥200 mg/L, and the distribution in each group was not diferent (p � 0.784).

Factors Associated with Cp Changes in the WD Cohort.
Te 352 follow-up records of 86 WD patients were grouped according to serum Cp levels (group A1 vs. A2, Figure 1) in order to identify factors associated with CP changes. Te clinical and biochemical fndings of groups A1 and A2 are shown in Table 2. During follow-up, serum Cp fuctuated in most patients according to the boundary value ≥140 mg/L: from high to low and back or not (n � 22 and 23, data not shown), always at a high level (n � 32), and only once recorded (n � 12). Linear regression analysis showed that serum copper, Sco, and Cp in 352 medical records had a signifcant linear correlation (R 2 � 0.769 and 0.775) (Figure S2). Te 24-hour urinary copper levels, which were basic and treated with DMPS, were signifcantly higher in group A1 ((both p ≤ 0.001), 137.2 (95.8, 213.1), and 1534.7 (1054.1, 2265.3)) than in group A2 (112.5 (82.7, 147.9) and 983.7 (523.7, 1618.9)). Group A1 signifcantly difered from group A2 in items such as AST (p � 0.011), c-GT (p ≤ 0.001), albumin (ALB) (p ≤ 0.001), and cholesterol (p � 0.017).

Factors Associated with WD and a Logistic Regression
Model. For the purpose of diferentiating WD patients with nearly normal serum Cp from those with other liver diseases, logistic regression analysis was performed between the training sets of groups A and B. Te characteristics of the training and validation sets of groups A and B are listed in Tables 3 and S2, respectively. In the training set, the serum Cp level of group A was signifcantly lower than that of the control group (group B, p � 0.002). Most of the liver metabolic indexes of the two groups had no signifcant diference, according to each factor of univariate binary logistic regression analysis, such as ALT, AST, c-GT, cholinesterase (CHE), ALB, TB, PT, INR, Child-Pugh scores, and proportion of LC, or Child-Pugh classifcation B and C (p � 0.075-0.748). Age (p ≤ 0.001), AP (p � 0.040), serum Cr (p ≤ 0.001), and cholesterol (Cho, p � 0.001) revealed a major diference between the two groups. Multivariate binary logistic regression analysis identifed 2 variables as independent predictors of WD: age and serum Cr (Table 4), with AUCs of 0.950 (95% CI: 0.907-0.993; p ≤ 0.001) in the training set and 0.903 (95% CI: 0.837-0.970; p ≤ 0.001) in the validation set (total set: 0.926, 95% CI: 0.887-0.965, and p ≤ 0.001) (Figure 2).
Te values of this regression model ranged from −13.51 to 11.26. Te cutof value was 0.617, with a sensitivity of 80.2%, specifcity of 92.9%, PPV of 90.8%, and NPV of 84.3% in the total cohort to identify WD patients (Table 5). If we use a cutof value of −1, in the total cohort, the sensitivity of diagnosing WD patients will rise from 80.2% to 90.7%, although its specifcity drops 12.3 points to 80.6%. Taking our purpose of reducing missed WD diagnoses into consideration, we thought that it was a reasonable and userfriendly number.

Discussion
We conducted a retrospective cohort study with the longest follow-up period (9.8 years). Te longitudinal study showed that the increased Cp in WD patients is related to the deterioration of liver function indexes and the rise of urinary copper excretion (Table 2). Ten, we conducted a casecontrol study and built a logistic regression model using two easily available markers: age and serum Cr to identify serum Cp-normalised WD from other hepatopathy patients.
Tis study found that, in WD patients, an increase in the serum Cp concentration mainly occurred when they presented liver function injury or were complicated with LC, such as abnormal changes in AST, c-GT, ALB, cholesterol, and internal copper load (Tables 1 and 2). Tis is consistent with previous research, in which the serum Cp concentration may be normal in hepatic WD with active liver diseases 12-14 and could decrease to low values after copper chelator treatment 15. By analysing the liver histology of 27 WD patients whose serum Cp levels varied from 210 to 269 mg/L, Peter Ferenci thought that it correlated with liver histologic activity (r � 0.47, p < 0.05) 15. We found a similar Cp change trend after treatment. At the same time, in our follow-up study, the serum Cp rose again after terminating the treatment. Tis may be explained by the fact that Cp is mainly produced by the liver and is an acute phase response protein to infection and infammatory agents [7,8]. Liver injury caused by the copper burden will increase Cp. Tis is likely why "normal" Cp is more common in WD patients with signifcant hepatic damage than in N-type WD patients (Tables 1 and 2) [15].
Cp is the main copper-binding glycoprotein in blood plasma, transporting 40-70% of total serum copper [8]. Serum Cp tested via immunological methods includes biologically inactive apo-Cp and biologically active holo-Cp binding with copper [28]. Sco represents the level of biologically active holo-Cp and has a greater diagnostic value than the Cp concentration determined by immunological techniques [26,28]. Terefore, we can easily understand International Journal of Clinical Practice that, in the longitudinal study, serum Cu and Sco had signifcant linear correlations with serum Cp (R 2 � 0.769 and 0.775, both p ≤ 0.001, Figure S2).
Te mutation distribution and allele frequency in the ATP7B gene from 64 WD patients in our study show that the hotspots are p.R778L (exon 8), p.P992L (exon 13), and p.R919G (exon 12) ( Figure S1), which are in accordance with Chinese prevailing hotspot mutations in WD patients [29]. Tere was no diference in mutation hotspots between the two group patients whose serum Cp was always higher than 140 mg/L or fuctuated (p � 0.332) (Table S1). Tese data did not show that the mutations of the ATP7B gene were associated with the change in Cp. Te same fndings have been found in North American, Russian, and Swedish samples, and their most common WD mutation (p.H1069E) had no signifcant correlation with ceruloplasmin activity [30].
Serum Cp and K-F rings are the most recognised and easily remembered tests for WD diagnosis. Unfortunately, they are always concealed and puzzling in hepatic-type WD patients 4-6. European guidelines and the Leipzig scoring system 6 have provided a perfect diagnostic pathway for WD diagnosis, but they rely on some parameters such as 24 h urinary copper excretion, ATP7B gene, or liver biopsy that are not readily available in ordinary clinical centers and are also expensive and time consuming. In our study, if we calculate the Leipzig score without urinary copper, ATP7B gene, and liver biopsy, the sensitivity to diagnose WD is only 64.0% (55/86 patients scored 3), which is obviously lower than that of our model, with a cutof value of 0.617 (80.2%). Based on the usual indicators, we developed a regression model to recognise these WD patients with "normal" serum Cp, which is helpful and handy.
Te parameters (age and serum Cr) adopted in this model are diferent from those used to identify WD from ALF patients in some other studies, such as AST/ALT and AP/TB [13,16,17,31]. Age and serum creatinine are associated with WD for the following possible reasons. (1) Age: Te majority of WD patients present between 5 and 35 years, and liver disease may precede neurologic manifestations by as much as 10 years [6]. In addition, the onset age of other causes of liver diseases is generally later, from 30 to 80 years, for instance, hepatitis B, steatosis hepatitis, and autoimmune hepatitis [32,33]. Hence, age is a crucial parameter for diferentiating WD from other liver diseases. (2) Cr: Nearly all (94%) serum Cr is derived from the healthy muscle and is the end product of creatine and creatine phosphate   International Journal of Clinical Practice     International Journal of Clinical Practice metabolism [34]. Creatine is synthesized primarily in the liver, and it is then transported to other tissues, particularly skeletal and heart muscles, to be phosphorylated. Te serum Cr is produced by a spontaneous, non-enzymatic anhydration of creatine in the muscle cells, which is pH and temperature-dependent. In healthy individuals, the kidney is the main route for Cr elimination. Cr has a low molecular weight, and it is not albumin-bound. Terefore, it is freely fltered at the level of the glomerulus [35]. In our study, the WD patients' serum Cr levels are lower than those in patients with other liver diseases. Tis diference could not be caused by impaired Cr production due to liver function damage, as Table 3 shows no signifcant diference in liver function between the two groups. In addition, none of the enrolled patients showed signs of kidney damage, so we believe that a decrease in serum Cr was not caused by kidney damage. Our study did not count muscle mass and 24 h urine Cr of the patients, so it is impossible to judge whether a decrease in Cr in WD patients is related to muscle mass [35,36]. Te accumulation of intracellular copper can induce oxidative stress and perturbing cellular function. Some studies showed that creatine may function as an antioxidant capable of reducing oxidative stress in skeletal myofbers, cardiomyocytes, and hepatocytes [37,38]. We hypothesize that excessive copper deposition in WD patients leads to overloaded oxidative stress and creatine depletion, which manifests as lower serum creatinine than in other liver diseases. However, whether Cr is directly related to oxidative stress remains unclear. Tis logistic regression model has a great AUC value (0.926, Figure 2). We do not insist that our model is perfect, and we recommend that it can only be used for preliminary screening, not defnitive diagnosis. Terefore, we selected two cutof values, one (0.617) with very high PPV and specifcity and the other (−1) with very high NPV and sensitivity (Table 5). Tis means that when you address a cutof value ≥0.617 to diagnose WD, 90.8% will be correct. Unfortunately, 19.8% WD patients will be missed, due to 80.2% of sensitivity. If we want to increase the sensitivity, −1 may be a better choice. It has a sensitivity of 90.7% and an NPV of 90.8%, which shows that 90.8% of the excluded patients are non-WD patients (Table 5). Overall, both 0.617 and −1 used as the cutof values for diagnosing and excluding WD have great predictive power. Nevertheless, we still need to be careful with a WD index from −1 to 0.617.
Limitations of our study include that this is a single center and retrospective study which may have bias and decreased the power of our results; the enrolled WD patients were relatively fewer due to it being a rare disease, and fewer patients had elevated serum Cp. In addition, because it is a retrospective study, we cannot obtain all patients' ATP7B genes. In fact, the clinical phenotypes, K-F ring, and increased copper excretion were sufcient for the diagnosis of WD. Finally, the pathological mechanism of reduced serum Cr in WD patients is still unclear.
In conclusion, WD patients' increased serum Cp concentrations were mostly pseudo and fnally returned to a reasonable level, corresponding to their liver improvement after treatment with a copper chelator, and were not signifcantly related to ATP7B gene mutations. A regression model was established with usual indicators, age and serum Cr, which have good performance in identifying suspicious WD patients from undefned patients whose serum Cp ≥ 140 mg/L. We hope that additional studies in patients who come from multiple centers could further verify this model. Wilson's disease.

Data Availability
Te data that support the fndings of this study are available from the corresponding author upon reasonable request.

Additional Points
What Is Known. (i) A decrease in serum ceruloplasmin (Cp) is a crucial marker for the diagnosis of Wilson's disease (WD). (ii) A few WD patients have nearly or completely normal serum Cp, which can lead to missed diagnosis. (iii) Tere are currently no routine and simple laboratory tests that can identify these WD patients with normalised serum Cp. What Is New. (i) Te normalisation of serum Cp in WD patients is related to their copper burden and liver function indexes, but not ATP7B gene mutation. (ii) Age and serum creatinine are independent risk factors for distinguishing WD from other liver diseases. (iii) Using a regression model built with age and serum creatinine in this study to distinguish undefned WD patients with serum Cp levels ≥140 mg/L has a positive predictive value of 89.3%.

Ethical Approval
Tis article does not contain any studies with human or animal subjects. All methods were carried out in accordance with relevant guidelines and regulations. Tis study approval and informed consent were exempted by the Clinical Research Ethics Committee of First Afliated Hospital, Zhejiang University School of Medicine.

Disclosure
Wenbin Hu and Bing Ruan contributed equally to this work and share last authorship. Te initial version of this manuscript was submitted as a preprint in the link https://www. researchsquare.com/article/rs-1081495/v1.

Conflicts of Interest
Te authors declare that they have no conficts of interest.