The Significance of Autoantibody Changes Over Time in Primary Biliary Cirrhosis

Abstract

Objectives

In primary biliary cirrhosis (PBC), the antimitochondrial antibody is a cornerstone of diagnosis, but there have been conflicting reports about the correlation of autoantibodies with disease stage and prognosis. We studied whether autoantibody levels changed over time and sought correlations with clinical outcomes in a cohort of patients with PBC.

Methods

We tested serial serum samples from patients with PBC at a research institution for several autoantibodies. Long-term clinical follow-up data were used to calculate the slopes (change over time) for autoantibodies, platelet count, Ishak fibrosis score, biopsy copper, and number of portal areas with bile ducts. An adverse clinical outcome was defined as hepatic decompensation, development of hepatocellular carcinoma, liver transplantation, or liver-related death. We performed linear or logistic regression or Fisher exact test as appropriate, adjusting for multiple comparisons.

Results

Twenty-seven patients with PBC with 145 serum samples were studied. Of the cohort, 85% was white, 81% was female, and median follow-up time was 20 years. Of the autoantibodies tested, only sp100 changed significantly over time. The sp100 slope was inversely associated with the Ishak fibrosis slope (parameter estimate, −0.05; P = .0003).

Conclusions

While changes in most autoantibodies over time do not seem to correlate with clinical outcomes in PBC, a change in the sp100 autoantibody level may have prognostic utility with respect to the development of fibrosis on liver biopsy.

Primary biliary cirrhosis (PBC) is a chronic liver disease characterized by progressive inflammatory destruction of small intrahepatic bile ducts, which ultimately leads to progressive hepatic fibrosis, cirrhosis, and end-stage liver disease. The pathogenesis of PBC remains unclear.

Characteristic if not diagnostic of PBC is the presence of several autoantibodies, particularly the antimitochondrial antibody (AMA). AMA was first linked to PBC 50 years ago when Doniach and Sherlock identified a reactivity in serum from patients with PBC that yielded a non–organ-specific cytoplasmic fluorescence in cells rich in mitochondria.¹ The antibody reactivity was later shown to be AMA and subsequently found to react with and inhibit the large enzymatic complex known as pyruvate dehydrogenase. The specificity of this antibody for PBC has been repeatedly shown, but the role of the autoantibody in the bile duct damage that characterizes this disease has remained unclear.² Furthermore, other autoantibodies, including antinuclear antibodies (ANAs), are found by indirect immunofluorescence in up to 50% of patients with PBC.^3,4 There have been conflicting reports about the correlation of AMA or other autoantibodies with disease stage and prognosis, with some data showing that there is no clinical utility in following antibody titers and others showing that AMA titers correlate with disease severity.^5-8 Two case series have found no significant change in AMA levels over time or with treatment.^6,9 In some reports, however, specific types of ANAs, such as the IgG3 isotype, anti-gp210, and anticentromere antibody, have been associated with a worse prognosis.^10-16

Although ursodeoxycholic acid is effective for most patients with PBC, others do not respond to our best studied treatment.¹⁷ Moreover, the progression of disease is largely unpredictable, and for that very reason, the various staging systems for PBC liver biopsy specimens is controversial. Aside from fibrosis progression, other histologic evidence of worsening disease includes copper accumulation and the loss of bile ducts. As in many forms of liver disease, the platelet count decreases as PBC progresses.¹⁸

In a proportion of patients treated at the National Institutes of Health (NIH) Clinical Center, patients have been found to become AMA negative and ANA positive during follow-up. However, these results are based on commercial assays that have changed over the past 20 years as our understanding of the immunobiology of PBC and our technical assays have improved. Accordingly, we have reassessed whether autoantibody levels change over time and whether the changes correlate with clinical manifestations, such as disease progression or worsening fibrosis or jaundice. Stored serum samples were tested by state-of-the-art assays and results were compared with clinical, laboratory, and histologic outcomes.

Materials and Methods

Database Assembly and Ethics Approval

We established a cohort of patients with PBC who had at least three serum samples stored over a period of at least 5 years while enrolled in studies of the natural history and therapy of PBC at the NIH Clinical Center. Because the patients had signed consent forms for studies on PBC in the past, the institutional review board granted a waiver of informed consent, and the current study was conducted as an amendment to an ongoing clinical research protocol (91-DK-0214: clinicaltrials.gov identifier NCT00001971), per the ethical guidelines of the 1975 Declaration of Helsinki. The NIH Biomedical Translational Research Information System, an existing database of patients with PBC, review of the medical record, and patient calls were used to collect data on demographic, laboratory, radiologic, histologic, and clinical outcomes.

Testing of Serum Samples

Serum samples obtained during protocol visits were stored at −80°C in the National Institute of Diabetes and Digestive and Kidney Diseases Clinical Core Facility, with minimal freeze-thaw cycles. Quality control measures included confirmatory immunoblotting, control sera from patients who had achieved sustained viral response after hepatitis C therapy, testing in duplicates, and blinding of laboratory researchers to duplicates. A material transfer agreement was obtained to transport samples between collaborating institutions.

Serum samples were coded and tested for the presence of IgG antibodies to MIT3 (state-of-the-art AMA consisting of a triple hybrid recombinant mitochondrial antigen), gp210, sp100, soluble liver antigen (SLA), chromatin, centromere, RNA Polymerase III, Scl-70, SS-A, SS-B, and SSA-52 by enzyme-linked immunosorbent assay (ELISA) (QUANTA Lite; Inova Diagnostics, San Diego, CA). The PBC screen is an ELISA detecting multiple mitochondrial and nuclear autoantibodies of IgG and IgM isotypes. Samples were also tested for ANA and AMA by indirect immunofluorescence assay on human epithelial cell substrate slides (Nova Lite HEp-2 ANA Kits; Inova Diagnostics). Serum samples were analyzed for specific immunoglobulin reactivity to mitochondrial proteins via conventional Western blotting methods and chemiluminescent detection (Pierce, Rockford, IL).

Statistical Analysis

The primary dependent variable was the change in platelet count over years of follow-up. Our primary independent variable was serial autoantibody levels. A power analysis was performed for testing the null hypothesis of no correlation between the two variables (correlation coefficient equal to zero). With a fixed sample size of 27 patients, we estimated 77% power in detecting a correlation coefficient of at least 0.5 at a significance level of .05.

Additional outcome variables of interest included biopsy Ishak fibrosis score, copper score, and percentage of portal areas with bile ducts present. A combined event was defined as hepatic decompensation (ascites, variceal bleeding, encephalopathy), development of hepatocellular carcinoma, liver transplantation, or liver-related death.

The data were analyzed with both cross-sectional and longitudinal methods. Cross-sectional analysis was performed to assess the relationship between initial autoantibody levels and initial outcome variables, as well as initial autoantibody levels and final outcome variables. Autoantibody levels were treated both as continuous variables and as categorical variables. Long-term follow-up data were used to determine the change over the study period (slopes) for the following measurements: autoantibodies, platelet count, Ishak fibrosis score, copper, and percentage of portal areas with bile ducts. The slopes are calculated as the measurements from first to last available value divided by the number of interceding years.

The comparison of continuous variables by group was performed using the t test, and the comparison of frequency variables was performed using the χ² test or Fisher exact test, as appropriate. General linear regression or logistic regression was used for multivariate regression analysis as appropriate, adjusting for potential covariates. We controlled the familywise error rate by adjusting the α via the Bonferroni correction for multiple comparisons. Analyses were performed with the statistical packages SAS (version 9.3; SAS Institute, Cary, NC), R (version 0.97.312; R Foundation for Statistical Computing, Vienna, Austria), and GraphPad Prism (version 5.0a; GraphPad Software, La Jolla, CA).

Results

Clinical Cohort

A total of 145 serum specimens from 27 patients with PBC were studied Table 1 . The cohort was largely non-Hispanic white (85.2%) and female (81.5%), and the median age was 50.4 years. Median follow-up time was 20 years (range, 6.0-30.5 years) and spanned from 1981 to 2013. After 1994, all patients were treated with ursodiol (600-900 mg/d). All patients underwent one liver biopsy, and 24 underwent two or more. Key variables for each of the 27 patients in the cohort are shown in Table 2 . The median alkaline phosphatase at the end of follow-up was 101 U/L (range, 51-363 U/L), and 13 patients experienced an improvement in their alkaline phosphatase over the course of the study. There was a decrease in median platelet count for the cohort from 284 × 10³/μL at the initial time point to 181 × 10³/μL at the last time point (P < .05).

Table 1. Clinical Characteristics of the Cohort^a.

Characteristic	Value
Age at first visit, median (range), y	50.4 (33.6-68.0)
Female sex	22 (81.5)
Race
White	23 (85.2)
Black	1 (3.7)
Other	3 (11.1)
Years of follow-up, median (range)	20.0 (6.0-30.5)
No. of biopsies, median (range)	6 (1-8)
No. of serum samples, median (range)	5 (3-8)

^aValues are presented as number (%) unless otherwise indicated.

Table 2. Changes in Autoantibody Levels and Clinical Variables Over Time for Each Patient.

Subject No.	Adverse Clinical Outcome	Follow-up Time, y	MIT3 Slope	sp100 Slope	Scl-70 Slope	RNA Polymerase III Slope	Platelet Slope	Fibrosis Slope	Copper Slope	PA With BD Slope
1	None	18.7	2.34	−0.89	0.07	0.08	5.71	0	0	−0.02
2	None	21.9	3.75	−0.01	−0.02	0.03	−8.48	−0.08	0	0.01
3	None	9.0	0	−1.66	−0.06	−0.07	−13.1
4	None	15.1	1.44	0.02	0.05	0.04	−8.66
5	None	9.7	−3.76	1.19	−0.01	0.11	−2.11	0		−0.06
6	None	21.4	−1.36	0.01	0.03	−0.05	−0.24	0	0	0.02
7	None	23.6	1.94	−2.76	−0.05	0.04	−1.1	0.17	0.09	0.005
8	None	20.0	1.99	0.51	0.03	0.18	0.19	0	0.16	−0.01
9	None	25.0	−0.3	−0.06	0.01	0.08	−6.65	0	0	0.04
10	None	24.6	1.61	−0.05	−0.12	−0.03	−6.5
11	None	11.5	1.81	−0.19	0.02	0.12	−3.15	−0.08	0	−0.0005
12	Hepatic decompensation	16.1	−2.18	0.2	−0.06	−0.05	−16.84			0.01
13	Liver transplantation	26.4	5.59	−1.36	0.06	0.24	−12.28	0.13	0.13	−0.03
14	None	21.8	−0.7	−1.02	−0.08	−0.02	−4.76	0	−0.06	0.01
15	None	17.9	1.41	−0.03	0.11	−0.02	−3.11	0	0.1	−0.05
16	None	23.8	1.39	0	0.26	0.01	−2.35	−0.12	0	0.01
17	HCC	30.5	0.21	−1.61	0.03	0.16	−8.98	0.13	0.08	−0.01
18	Liver−related death	21.8	−0.59	−0.08	0.1	0.66	−1.15	−0.09	0.2	0.003
19	None	6.0	−6.21	−0.02	−0.27	−0.25	−24.51
20	None	14.7	3.32	0.4	0.005	−0.2	−13.68	0.15	0.07	−0.03
21	None	15.1	−1.44	−10.27	−0.05	−0.06	−3.92
22	None	20.5	−2.3	−0.21	−0.06	−0.12	−3.63	0.15	−0.19	−0.06
23	None	26.7	−0.41	−0.04	−0.05	−0.05	−4.44	−0.05	−0.07	0.01
24	None	19.0	−0.69	0.004	0.09	−0.02	−14.81	0.2	0.1	−0.02
25	None	18.7	−0.7	0.03	0.01	0.02	−10.25	−0.13	0.13	−0.05
26	None	14.8	3.27	0.08	−0.1	−0.04	−0.07
27	None	20.7	0.98	−7.49	−0.03	0.04	−6.56	0.42	0.11	−0.04

BD, bile duct; HCC, hepatocellular carcinoma; PA, portal area.

Autoantibody Testing

The 145 serum samples from patients with PBC had been in storage for variable lengths of time (range, 3-252 months). Initial serum samples were positive for the PBC screen in 26 (96%) of 27 patients. Furthermore, MIT3 was present in the initial serum sample in 25 (93%) of 27 patients. Initial serum samples were positive for SSA-52 in 13 (48%) patients, sp100 in six (22%) patients, and gp210 in six (22%) patients. Five (19%) patients were initially positive for centromere and four (15%) patients for SS-A. One (4%) patient was initially positive for chromatin, and one (4%) was initially positive for SS-B. None of the initial serum samples were positive for RNA Polymerase III, SLA, or Scl-70. All patients were positive for ANA at onset and throughout follow-up. All patients experienced changes in their ANA immunofluorescence pattern over time.

Of the autoantibody levels tested, most did not change significantly over time (Supplemental Table 1; supplemental material can be found at http://bit.ly/TanaOct15). MIT3 titers did not change significantly over time, with an initial median value of 129.9 and final median value of 132.4 (Supplemental Figure 1). All patients, except for one, were stably MIT3 positive or negative on all serum samples tested. Most (15/27 [56%]) patients had sp100 levels that remained approximately stable over time (Supplemental Figure 2).

Histologic Outcomes

There was little histologic evidence of disease progression despite the fact that the interval between the first and last biopsies ranged from 6.5 to 21.5 years. The proportion of portal areas with bile ducts did not change over time (median initial 70%, median final 61%). Scores for copper retention did not change on average (median initial score 1, median final score 1). The cohort's Ishak fibrosis scores changed minimally during follow-up (median initial 2, median final 2). None of the patients had cirrhosis on their initial biopsy specimen, and the final biopsy specimen showed cirrhosis in only two patients. Seven patients experienced an increase in fibrosis by one or more Ishak fibrosis stages over the course of follow-up. Six patients experienced a decrease in their fibrosis scores by one or more Ishak fibrosis stages.

Clinical Outcomes

Four patients experienced an adverse clinical outcome during the follow-up period. One patient experienced variceal bleeding but lived, one patient underwent liver transplantation, one patient died of end-stage liver disease, and one patient developed hepatocellular carcinoma. The mean time from initial visit to adverse clinical outcome was 16.4 years (range, 2.9-30.5 years). The patients with an adverse clinical outcome did not differ significantly from those without in terms of initial alkaline phosphatase, alanine aminotransferase, total bilirubin, IgM, prothrombin time, biopsy fibrosis, or biopsy portal areas containing bile ducts.

Correlation Between Autoantibodies and Outcomes

On longitudinal data analysis, there was no correlation between changes in autoantibody levels (PBC screen, MIT3, sp100, gp210, RNA Polymerase III, SLA, Scl-70, chromatin, centromere, SSA-52, SS-A, SS-B, or ANA) and changes in outcome variables (platelets, histologic features, or clinical outcomes), with one exception. The sp100 slope was inversely associated with the Ishak fibrosis slope on linear correlation analysis, with a parameter estimate of −0.05, and P = .0003 Figure 1 . There was not a significant relationship between the MIT3 slope and any of the outcomes. We adjusted for multiple comparisons using a Bonferroni method. However, if a less stringent P value threshold of .05 were used, the following relationships would also be statistically significant: the association between Scl-70 slope and platelet slope, the association between RNA Polymerase III slope and platelet slope, and the relationship between RNAP Polymerase III slope and biopsy copper slope.

Figure 1.

The sp100 slope is inversely associated with the Ishak fibrosis slope (P = .0003).

In cross-sectional analysis adjusting for multiple comparisons, initial autoantibody levels were neither associated with disease severity at onset nor predictive of future outcomes. However, several relationships had P values less than .05, and these are shown in Table 3 . Of note, initial anti-La/SS-B autoantibody levels were associated with initial platelet counts, final platelet counts, and final Ishak fibrosis scores (P = .03, P = .03, and P = .05, respectively).

Table 3. Relationships Between Initial Autoantibody Levels and Outcome Variables With P Values That Are Less Than .05.

Relationship	P Value
Initial SS-B and initial platelet count	.0272
Initial RNA Polymerase III and initial biopsy copper	.0149
Initial SS-A and final platelet count	.0331
Initial SS-B and final platelet count	.0311
Initial SSA-52 and final Ishak fibrosis	.0504
Initial SS-A and final Ishak fibrosis	.0127
Categorical initial SSA-52 and final platelet count	.0289
Categorical initial sp100 and final Ishak fibrosis	.0313
Categorical initial SS-A and final Ishak fibrosis	.0127
Categorical initial SS-B and final Ishak fibrosis	.0447

Discussion

Our clinical experience at the NIH Clinical Center suggested that some patients experience changes in autoantibody results with treatment. We studied autoantibody levels in serial serum samples from a cohort of well-characterized patients with PBC to test this hypothesis. Of the host of other autoantibodies we examined, sp100 stood out as changing significantly over time. Furthermore, there was a significant inverse relationship between change in sp100 autoantibody level over time and change in fibrosis score on serial liver biopsy specimens. Our results suggest that while MIT3 does not change much in most patients, it does change somewhat in a few patients. Two patients demonstrated an inverse relationship between MIT3 slope and fibrosis slope, similar to the inverse relationship we found between sp100 slope and fibrosis slope. In cross-sectional analysis, the initial anti-La/ SS-B autoantibody trended toward a significant relationship with final platelet counts and fibrosis scores.

Although the role of AMAs in the pathogenesis of PBC has not been entirely elucidated, there is evidence that AMAs have a functional role in the pathogenesis of PBC.¹⁹ Therefore, it was reasonable to think that as the disease responds to ursodiol or other therapy, AMA levels might fluctuate in parallel with disease activity. Somewhat surprisingly, this was not substantiated by the current study.

In this study, we examined not only AMA levels but also levels of specific ANAs. sp100 is a nuclear antigen, consisting of at least three autoantigenic domains. A recent meta-analysis of 21 studies concluded that sp100 has a specificity of 98% for the diagnosis of PBC.²⁰ In one study of 70 patients with PBC, the sp100-positive patients were found to have higher alkaline phosphatase values, γ-glutamyltransferase, and IgM levels.²¹ Another study of 170 patients with PBC found that sp100 levels correlated with the Mayo risk score.²² A longitudinal study of 110 patients found that declining sp100 titers correlated with improvement in the Mayo risk score and response to ursodiol.²³ Therefore, our finding that the change in sp100 level is associated with changes in fibrosis builds on prior work but is paradoxical in that other studies associated increasing sp100 levels with worse clinical outcomes. Some of this discrepancy may be due to limited power with our number of patients or the different outcomes studied (ie, decompensation vs histologic fibrosis), but it may be that changes in autoantibody levels, whether higher or lower, signify a physiologic shift with progressive disease. While any instability of the tested autoantibodies in stored serum would affect our results, our laboratory has previously performed nested studies on serum samples stored for over a decade with similar results for AMA determination. A limitation of our study is that not all patients had paired biopsies, and liver biopsy in general is subject to sampling error, perhaps a larger problem in PBC than in other chronic liver diseases.

Nevertheless, the longitudinal nature of our data (ie, long-term follow up of serum, histology, and clinical data) represents a valuable contribution to the field. These findings leave us asking why most autoantibodies, some of which are so helpful diagnostically, have no clinical significance in the course and outcome of the disease. It has been shown that the presence of AMAs precedes the clinical manifestations of the disease and oftentimes diagnosis by many years.^24,25 Some would argue that just as antibody responses to vaccination are generally stable over many years, likewise we would not expect these AMA levels to change dramatically over time.

Finally, it is interesting to compare our findings in patients with PBC with what is known about the significance of autoantibody levels in other autoimmune diseases. In myasthenia gravis, titers of acetylcholine receptor antibody do not correlate with the severity of disease.²⁶ Furthermore, although antibody levels decrease in almost all patients whose disease improves, they also fall in patients who now experience clinical improvement.²⁷ In rheumatoid arthritis, the presence of rheumatoid factor and anti–citrullinated peptide antibodies is associated with a higher risk of erosive joint damage and functional impairment. High titers of rheumatoid factor seem to predict a more severe disease course with extra-articular manifestations, a higher rate of erosion formation, and worse radiologic outcomes.^28,29 Although anti–double-stranded DNA antibodies are highly specific for systemic lupus erythematosus, the titer does not correlate with clinical severity or course for many patients.^30,31 It seems that, for the most part, autoantibody levels do not correlate with clinical severity in PBC and other autoimmune diseases, perhaps because qualitative characteristics of the autoantibodies trump their quantitative level.

Conclusions

Our careful analysis of serum samples from 27 patients with PBC followed closely over several years suggests that the sp100 autoantibody can change significantly over time. This change was inversely related to changes in biopsy fibrosis score, suggesting that changes in sp100 level despite treatment might raise a clinician's suspicion for progressive disease. Prospective studies on larger numbers of patients will determine if this association holds true.