Diagnostic value and utility of the simplified International Autoimmune Hepatitis Group (IAIHG) criteria in acute and chronic liver disease

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Diagnostic criteria for autoimmune hepatitis (AIH) have been created and revised by the International Autoimmune Hepatitis Group (IAIHG). Simplified criteria have been created, but remain independently unvalidated. We report on the diagnostic accuracy of the simplified criteria in patients across a range of diagnoses, including a subset of patients presenting with fulminant liver failure who required liver transplant. Patients with AIH and non-AIH etiologies of liver disease were identified from dedicated patient databases. Parameters relevant to the simplified and 1999 IAIHG criteria were recorded, and sensitivity, specificity, and positive and negative predictive values for scores of ≥6 (probable AIH) and ≥7 (definite AIH) were calculated. A total of 549 patients with chronic liver disease were evaluated, (221 with AIH, 26 with variant syndromes, and 302 with non-AIH). For scores ≥6, sensitivity was 90%, and specificity was 98% with positive and negative predictive values of 97% and 92%, respectively. For scores ≥7; sensitivity was 70%, and specificity was 100% with positive and negative predictive values of 100% and 74%, respectively. Seven false positive diagnoses of AIH occurred, all with simplified scores of 6. Concordance with 1999 criteria was 90% for probable and 61% for definite AIH. The frequency of an overall diagnosis of AIH (probable or definite AIH) among the 70 patients with fulminant liver failure was 24% for simplified criteria and 40% for 1999 criteria, respectively. Conclusion: The simplified criteria retain high specificity but exhibit lower sensitivity for scores of ≥7. The explanations for this are unclear but may relate to loss of such discriminating information as response to corticosteroids. Prospective evaluation of these criteria is required to corroborate these observations. (HEPATOLOGY 2009.)

The International Autoimmune Hepatitis Group (IAIHG) was conceived in 1993, consisting of experts in the field of autoimmune liver disease. Convened originally to develop criteria to aid in the diagnosis of autoimmune hepatitis (AIH) that would be useful in clinical practice (distinguishing chronic active hepatitis related to hepatitis C virus [HCV] infection from that of a true autoimmune etiology) and to allow comparison between patients in trials of AIH, these criteria were widely accepted and incorporated into clinical practice in both nontransplant1–5 and transplant settings.6, 7

The initial deliberations led to diagnostic criteria that defined individuals into one of three categories: not AIH, probable AIH, and definite AIH.1 Although cumbersome, when applied, these criteria could be used to diagnose AIH with a high degree of sensitivity, albeit with a lower specificity. Prospective validation of these cohorts suggested a sensitivity for the diagnosis of AIH of between 97% and 100%.2–4 Over time, it became evident that patients with other autoimmune liver conditions such as primary sclerosing cholangitis (PSC) and primary biliary cirrhosis (PBC) could be ascribed a false positive diagnosis of AIH. Therefore, a desire arose, not just to give the criteria greater specificity, but also to allow easier implementation in clinical, rather than research, practice. This led to a revision of the original criteria in 1999.5 These revised criteria proposed modification of the scores allocated to the presence, or otherwise, of cholestatic liver function tests, drug history, liver histology, and response to treatment. The original revised, and widely applied, IAIHG criteria for the diagnosis of AIH are summarized in Table 1. The revised score, although a useful adjunct to the diagnosis of AIH in cases with atypical features, remained cumbersome due to the sheer number of variables (n = 12) analyzed, which led to criticism of its application. Furthermore, the diagnosis of AIH continued to be challenging, especially in patients with atypical histological, biochemical, and immunological features.

Table 1. Original Revised (1999) Criteria for the Diagnosis of AIH
Parameter/DiscriminatorScore
  1. Adapted from Alvarez et al.5

Female sex+2
ALP:AST (or ALT) ratio 
 <1.5+2
 1.5–3.00
 >3.0−2
Serum globulins or IgG above normal 
 >2.0+3
 1.5–2.0+2
 1.0–1.5+1
 <1.00
ANA, SMA, or LKM-1 
 >1:80+3
 1:80+2
 1:40+1
 <1:400
AMA-positive−4
Hepatitis viral markers 
 Positive−3
 Negative+3
Drug history 
 Positive−4
 Negative+1
Average alcohol intake 
 <25 g/day+2
 >60 g/day−2
Liver histology 
 Interface hepatitis+3
 Predominantly lymphoplasmacytic infiltrate+2
 Rosetting of liver cells+1
 None of the above−5
 Biliary changes−3
 Atypical features−3
Other autoimmune disease(s) in either patient or first-degree relative+2
Optional additional parameters 
 Seropositivity for other defined antibodies+2
 HLA DR3 or DR4+1
Response to therapy 
 Remission alone+2
 Remission with relapse+3
Interpretation of aggregate scores 
 Pretreatment: 
  Definite AIH>15
  Probable AIH10–15
 Posttreatment: 
  Definite AIH>17
  Probable AIH12–17

Consequently, the IAIHG group have recently streamlined the diagnostic criteria, with a sharp focus on the clinical diagnosis of AIH, and the aim of making them readily usable “at the bedside”.8 As a result, the new, simplified criteria incorporates only four parameters: presence of associated autoantibodies, immunoglobulin G (IgG) level, liver histology, and the absence of viral hepatitis. The reduction in number of parameters analyzed was based on the outcomes of stepwise logistic regression which identified the above features, when combined, as having the greatest predictive value for the diagnosis of AIH with an area under the curve (AUC) on receiver operating characteristic (ROC) of 0.99. The simplified IAIHG criteria for the diagnosis of AIH are summarized in Table 2.

Table 2. Simplified Diagnostic Criteria for the Diagnosis of AIH
ParameterDiscriminatorScore
  1. Adapted from Hennes et al.8

  2. ≥6 points: Probable AIH

  3. ≥7 points: Definite AIH

ANA or SMA +≥1:40+1
ANA or SMA +≥1:80 or+2
LKM +≥1:40 or+2
SLAPositive+2
IgG Level>Upper limit of normal+1
 >1.1× Upper limit of normal+2
Liver HistologyCompatible with AIH+1
 Typical of AIH+2
Absence of Viral HepatitisNo0
 Yes+2

Currently, in this new system, a score of ≥6 (8 being maximum) equates to “probable AIH” whereas a score of ≥7 or above denotes “definite AIH”. The specificity of this score in the training and validation sets obtained from retrospective analysis of a multinational cohort is 97% for a probable diagnosis and 99% for a definite diagnosis of AIH. Meanwhile, sensitivity was reported as 88% for probable and 81% for a definite diagnosis of AIH using these criteria.

What remains uncertain, however, is whether these criteria remain reproducible outside of the IAIHG multinational patient cohort, although a recent publication has described performance characteristics of the score in a series which contributed to the cohort used in the development of the simplified criteria.9 We therefore set out to examine the validity and utility of the new simplified IAIHG diagnostic criteria and make an assessment of its sensitivity and specificity in the diagnosis of AIH both in comparison to the IAIHG simplified criteria and also to the original, revised criteria published in 1999.

In addition, it is not known how many patients (if any) in the simplified IAIHG criteria presented with fulminant hepatic failure (FHF) and, although this group represents a relatively infrequent presentation of AIH, these patients present difficult challenges diagnostically. Therefore, it is of interest to assess the utility of the simplified criteria in this patient group. Indeed, work from our own institution, among others, suggests that features of autoimmunity (such as autoantibodies) are a frequent finding in patients with FHF despite the absence of classical clinical features of AIH.10–13

Consequently, the utility of the simplified diagnostic criteria in this setting has not been established. Moreover, if it was possible to aid the early identification of an autoimmune etiology of FHF, prompt treatment might improve clinical outcome and prevent the need for liver transplantation.

Abbreviations

AIH, autoimmune hepatitis; ALD, alcoholic liver disease; AUC, area under the curve; FHF, fulminant hepatic failure; HBV, hepatitis B virus; HCV, hepatitis C virus; IAIHG, International Autoimmune Hepatitis Group; IgG, immunoglobulin G; NASH, nonalcoholic steatohepatitis; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; ROC, receiver operating characteristic.

Patients and Methods

Patients with AIH were identified from a dedicated database, collected prospectively since 1971, and continuously updated. Patients with other liver diagnoses (hepatitis B virus [HBV] and HCV infection, PSC, PBC, hemochromatosis, Wilson's disease, drug-induced liver injury, nonalcoholic steatohepatitis [NASH], alcoholic liver disease [ALD], PBC/AIH overlap [variant], and PSC/AIH overlap [variant] syndromes) were also retrospectively identified from the Institute of Liver Studies, Liver Database. All diagnoses were made via standard clinical criteria and, for drug-induced liver injury, by exclusion of all other causes in patients exposed to known hepatotoxic agents.

For the cohort with non–acetaminophen induced liver failure, individuals were retrospectively identified from a dedicated database of patients who underwent transplantation at King's College Hospital, National Health Service Foundation Trust, London, UK. Individuals who underwent transplantation were chosen because the majority with this mode of presentation do not have a liver biopsy before transplant due to the risk of hemorrhage. Moreover, because the simplified criteria mandate biopsy material, patients who underwent transplantation for FHF have explant histological material available for analysis in all instances.

Only patients with complete demographic, laboratory, and clinical data pertinent to both the simplified and original revised criteria, at the time of their initial diagnosis, were included in this study. In addition, histology was available in every patient. Viral hepatitis was excluded via second-generation and third-generation microparticle enzyme immunoassays for HBV and HCV. For the purposes of this study, overlap syndromes were considered as AIH, as recently reported by Hennes et al. in their description of the simplified criteria for the diagnosis of AIH.8 Scores derived from the new, simplified criteria and the 1999 IAIHG criteria scores were then calculated. The sensitivity and specificity of these scores for a probable or definite diagnosis of AIH were then calculated, and the positive and negative predictive values were recorded.

For patients with FHF, only the relative frequency with which a probable or definite diagnosis of AIH was determined by each score was recorded. It is not possible to calculate the relative sensitivity or specificity for either the 1999 or simplified diagnostic criteria in this clinical context, because it is frequently impossible to ascribe a firm diagnosis in these patients. Diagnosis in these situations is rendered difficult both due to the severity of hepatic insult at presentation (rendering histological evaluation extremely difficult), as well as the immunoparesis commonly seen in the critically ill patient in whom both autoantibodies and/or elevated IgG concentrations may be absent.14 It is, however, of interest to compare the simplified criteria scores for this patient cohort with the original revised criteria (as the “gold standard”) scores in order to document the relative frequencies with which these criteria identify individuals with either a probable or definite diagnosis of AIH.

Statistical Analysis.

All results are displayed as percentages, or, for continuous variables, median and range. Comparisons of proportions were analyzed by chi-squared test or Fisher's exact test if any of the variables equaled less than 5. ROC curves were generated to analyze and compare the accuracy of the simplified and original revised criteria in the diagnosis of AIH. All statistical analysis was performed using the SPSS statistical software package version 14 (SPSS Inc., Chicago, IL).

Results

A total of 549 patients, comprising 348 females (63%) and 201 males (37%), were identified who had the relevant clinical, laboratory, and histological information available at the time of diagnosis in order to calculate both the 1999 and the simplified IAIHG criteria scores. Of these, 221 patients held a diagnosis of AIH (79% female) with the remaining 328 (53% female) patients having diagnoses of HBV, HCV, PBC, PSC, NASH, AIH/PBC overlap, AIH/PSC overlap, ALD, Wilson's disease, alcohol-related liver disease, hemochromatosis, and drug-induced liver injury. Drug-induced liver injury was related to statins in two patients, nandrolone in two, amoxicillin/clavulanic acid in two, methylene-dioxy-N-methamphetamine (MDMA) in one, synthetic estrogens in one, and enalapril in another. The laboratory and demographic features, in addition to the median simplified and 1999 IAIHG scores by underlying liver diagnosis, are summarized in Table 3.

Table 3. Demographic and Laboratory Features with 1999 and Simplified Criteria Scores Across Diagnoses in the Validation Cohort
Liver Diagnosis (n)F:M (%)AST Median (range)AST/Alp Ratio Median (range)Autoantibodies(%)IgG Median (range)1999 ScoreSimplified Score
  • Antinuclear antibody, anti–smooth muscle, or anti–liver-kidney microsomal antibody titers ≥1:40.

  • AIH, autoimmune hepatitis; ALP, alkaline phosphatase; AST, aspartate aminotransferase; F, female; FHF, fulminant hepatic failure; HBV, hepatitis B virus; HCV, hepatitis C virus; IgG, immunoglobulin G; M, male; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis.

AIH (221)79:21617 (23–4603)0.13 (0.01–3.9)8424.8 (4.7–70.7)22 (10–28)7 (3–8)
HBV (51)27:7344 (19–1507)0.54 (0.03–3.4)1016.3 (13–23.6)0 (0–3)0 (0–4)
HCV (53)36:6456 (23–210)0.53 (0.10–2.6)2613.8 (9.4–25.4)0 (0–8)1 (0–4)
NASH (49)57:4346 (21–937)0.77 (0.17–4.2)1411.4 (7.6–29.4)5 (1–13)2 (2–5)
PBC (57)96:494 (28–537)1.60 (0.60–6.7)1915.3 (6.9–28.7)1 (0–13)2 (1–6)
PSC (49)47:5382 (31–883)1.60 (0.33–4.8)2018.8 (7.4–38.0)4 (0–10)3 (2–6)
PBC/AIH variant (9)100:0131 (44–1200)0.35 (0.04–2.6)8918.1 (9.6–26.9)14 (7–17)5 (4–8)
PSC/AIH variant (17)35:65202 (63–1750)0.86 (0.10–3.5)9421.8 (14.0–50.4)13 (9–21)6 (0–8)
Alcohol-related (20)55:4562 (27–217)0.80 (0.2–2.4)2017.8 (11.4–32.8)1 (0–9)3 (2–6)
Hemochromatosis (7)14:8651 (26–180)0.72 (0.2–1.6)015.9 (12–25)0 (0–3)2 (2–4)
Drug-induced (9)44:56111 (24–13078)0.56 (0.01–2.1)1114 (8.0–22.4)1 (0–7)2 (2–5)
Wilson's Disease (7)43:5780 (32–347)1.00 (0.06–4.9)020.2 (7.6–23.0)5 (1–7)3 (2–4)
FHF (70)74:26958 (66–13874)0.06 (0.001–0.9)3711.8 (1.9–63.9)7 (0–18)3 (2–8)

Patients with AIH.

The median age at diagnosis of patients with AIH was 46 years (range 5-80).

The simplified criteria identified 132 (60%) patients as definite AIH, 73 (33%) patients as probable AIH, and 16 (7%) as not having AIH, in contrast with the 1999 criteria, which for this cohort, identified 220 of 221 (99.5%) patients with definite AIH and only one as probable AIH (1 of 221 = 0.5%). The sensitivity of the simplified criteria in this cohort was 90% for a probable diagnosis of AIH and 70% for a definite diagnosis. Meanwhile, the specificity was 98% and 100% for probable and definite AIH, respectively. The positive and negative predictive values were 97% and 92% for a probable diagnosis and 100% and 74%, respectively, for a definite diagnosis of AIH.

In comparison, evaluation of the 1999 IAIHG criteria demonstrated sensitivities and specificities of 100% and 97% for probable AIH and 99.5% and 98% for definite AIH, respectively. These results are summarized in Table 4.

Table 4. Sensitivity, Specificity, Positive Predictive Value (PPV), and Negative Predictive Value (NPV) of the Simplified and 1999 IAIHG Criteria for the Diagnosis of AIH
CriteriaSensitivitySpecificityPPVNPV
Simplified Criteria    
 Probable diagnosis AIH (6–7)90%98%97%92%
 Definite diagnosis AIH (≥7)70%100%100%74%
 Overall Diagnosis AIH (≥6)90%98%97%92%
1999 Criteria    
 Probable diagnosis AIH (10–15)100%97%96%100%
 Definite diagnosis AIH (≥15)99%98%97%99%
 Overall Diagnosis AIH (≥10)100%97%97%99%

Concordance between the original revised and the simplified scores for a probable or definite diagnosis of AIH was 90% and 61%, respectively. An overall diagnosis of AIH (probable or definite) was recorded in 100% of patients using the 1999 criteria but only in 90% using the simplified criteria. In addition, ROC curves for an overall diagnosis of AIH demonstrated an AUC of 0.934 (95% confidence interval, 9.09-9.59) for the simplified criteria whereas the 1999 criteria had a greater AUC of 0.979 (95% confidence interval, 0.966-0.993) (Fig. 1).

Figure 1.

Receiver operating characteristic (ROC) curves for the simplified and 1999 criteria in the diagnosis of AIH.

Variant Syndromes.

Of the 26 patients with a variant syndrome, nine had PBC/AIH and 17 had PSC/AIH. All 26 were treated with prednisolone ± azathioprine following their diagnosis. Fifteen (58%) were ascribed a diagnosis of AIH (seven probable, eight definite) by the simplified criteria as compared to 24 (92%) patients (15 probable and nine definite) by the 1999 criteria, but this did not reach statistical significance (P = 0.17). Patients with the PSC/AIH variant were more likely to be ascribed a probable or definite diagnosis of AIH on the simplified criteria than those with the PBC/AIH variant, although this again did not reach statistical significance (65% versus 44% P = 0.42). However, comparison of patients with pure AIH in contrast to those with variant AIH revealed that individuals with pure AIH were significantly more likely to be female (79% versus 58%; P = 0.013), icteric at presentation (69% versus 23%; P = <0.0001), with a higher bilirubin (59 μmol/L versus 17 μmol/L; P = <0.0001), aspartate aminotransferase (AST) (617.5 IU/L versus 127.5 IU/L; P = <0.0001) and international normalized ratio (1.21 versus 0.98; P = <0.0001) at diagnosis. Patients with pure AIH also had higher median simplified criteria biopsy scores (2 versus 1 < 0.0001) and were more likely to respond to corticosteroids (95% versus 73% P = 0.002). Median IgG concentrations trended toward being higher in patients with pure AIH (24.8 g/L versus 19.7 g/L) but did not reach significance (P = 0.16), whereas the proportion of patients with positive autoantibodies in titers of either ≥1:40 or ≥1:80 did not significantly differ between the two groups. These results are summarized in Table 5.

Table 5. Comparison of Demographic, Laboratory, and Clinical Variables at Presentation Between Patients Clinically Diagnosed with Either Pure or Variant AIH in the Study Cohort
VariablePure AIH n = 221Variant AIH n = 26P Value
  1. ALP, alkaline phosphatase; AST, aspartate aminotransferase; F, female; IgG, immunoglobulin G; INR, international normalized ratio; M, male.

F:M79:2158:420.013
Icteric69%23%<0.0001
Bilirubin59 (7–1096)17 (4–246)<0.0001
AST617.5 (23–4603)156.5 (44–1750)0.0001
ALP:AST ratio0.14 (0.01–3.80)0.794 (0.03–3.5)<0.0001
INR1.21 (0.87–2.80)0.98 (0.86–1.60)<0.0001
Cirrhotic49%38%0.31
Exclusion of viral hepatitis100%100%0.99
Simplified criteria biopsy score2 (0–2)1 (0–2)<0.0001
Responders to treatment95%73%0.002
IgG (g/dL)24.8 (4.7–70.7)19.7 (9.6–50.4)0.163
Autoantibodies   
 ≥1:4081%92%0.14
 ≥1:8067%73%0.56

Patients with a False Negative Diagnosis of AIH.

Twenty-seven patients (11%) with a clinical diagnosis of AIH or variant AIH were not ascribed a diagnosis of AIH via utilization of the simplified criteria. Having identified significant clinical differences between patients with pure and variant AIH, analysis focused on the majority of patients who have pure AIH. Sixteen patients (7.2%) with pure AIH were not ascribed a diagnosis of AIH on the simplified criteria, whereas the 1999 criteria identified all these as having either probable or definite AIH. When comparing patients ascribed a false negative or a true positive diagnosis of pure AIH, there was no statistical difference between the ratio of females to males, frequency of an icteric presentation, bilirubin, AST, and alkaline phosphatase/AST ratio at presentation. The major differences therefore related to lower concentrations of IgG (median 12.5 versus 25.5 g/dL; P = <0.0001) and proportion of autoantibody titers of ≥1:40 (38% versus 84%; P = <0.0001) or ≥1:80 (6% versus 72%; P = <0.0001) among false negative cases. The demographic, laboratory, and clinical characteristics of patients with a false negative or true positive diagnosis of AIH is summarized in Table 6.

Table 6. Comparison of Demographic, Laboratory, and Clinical Variables at Presentation Between False Negative and True Positive Cases of Pure AIH as Ascribed by the Simplified Diagnostic Criteria
VariableFalse Negatives n = 16True Positives n = 205P Value
  1. ALP, alkaline phosphatase; AST, aspartate aminotransferase; F, female; IgG, immunoglobulin G; INR, international normalized ratio; M, male.

F:M81:1979:210.99
Icteric81%68%0.40
Bilirubin97 (11–487)58 (7–1096)0.60
AST612 (41–2060)619 (23–4603)0.59
ALP:AST ratio0.11 (0.04–0.75)0.16 (0.01–3.80)0.84
INR1.3 (0.94–2.80)1.2 (0.87–2.80)0.53
Cirrhotic38%50%0.34
Exclusion of viral hepatitis100%100%0.99
Simplified criteria biopsy score2 (0–2)2 (0–2)0.28
Responders to treatment100%94%0.99
IgG (g/dL)12.5 (8.00–20.00)25.5 (4.72–70.70)<0.0001
Autoantibodies   
 ≥1:4038%84%<0.0001
 ≥1:806%72%<0.0001

In view of the reduction in sensitivity for a diagnosis of AIH recorded via the utilization of the simplified criteria, we hypothesized that the addition of responsiveness to corticosteroids might alter the test performance characteristics. We nominally awarded a score of +2 points for a complete response and 0 for no response, and used revised cutoffs of ≥7 for probable and ≥8 for a definite diagnosis of AIH. This resulted in improved sensitivity for this cohort from 90% to 96% for a probable diagnosis, and from 70% to 92% for a definite diagnosis of AIH. Subsequently, the sensitivity for an overall diagnosis of AIH increased from 90% to 96% (P = <0.0001).

Patients with a False Positive Diagnosis of AIH.

Utilizing the simplified criteria, there were seven false positives out of 328 (2.1%) non-AIH patients ascribed a probable diagnosis, and none with a definite diagnosis of AIH. In comparison, the original revised IAIHG criteria ascribed 10 of 328 (3.0%) patients as having probable AIH, and none as having definite AIH. Using the simplified criteria, the seven false positive cases defined as probable AIH comprised four patients with PSC, two with PBC, and one with ALD. The false positive, probable AIH cases, identified by utilization of the original, revised criteria carried diagnoses of NASH in six patients, lone PBC in two patients, and lone PSC in two patients.

Patients with Fulminant Hepatic Failure.

Seventy patients comprising 52 females (74%) and 18 males (26%) with non–acetaminophen induced FHF had the relevant clinical, laboratory, and histological data available to calculate the original revised and simplified criteria scores. In this cohort, the frequency of a diagnosis of AIH, based on the simplified criteria, was 10% for probable (seven patients) and 14% (10 patients) for definite AIH, whereas on the 1999 criteria this equated to 31% (22 patients) and 9% (six patients), respectively. The frequency of an overall diagnosis of AIH was therefore 24% on the simplified criteria and 40% on the 1999 criteria. Only one patient in the seronegative group met criteria for probable AIH, and none met the criteria for definite AIH, via the simplified criteria whereas nine patients met the criteria for probable AIH and none for definite AIH via the 1999 criteria.

Prior to the application of either of the 1999 or simplified criteria, 28 of 70 (40%) patients were diagnosed with AIH on clinical grounds, and 42 (60%) were deemed “seronegative” liver failure. In those diagnosed with AIH by clinical perception, 50% were identified as such by the simplified criteria versus 60% for the 1999 criteria. Concordance between the 1999 and simplified criteria for an overall diagnosis of AIH was 13 of 28 (46%) patients. In four patients, the simplified criteria suggested a definite diagnosis of AIH in whom the 1999 criteria would have graded them as only probable. Additionally, there were 11 instances (16%) in which the 1999 criteria attributed a probable diagnosis of AIH, but the simplified score categorized these individuals as not having AIH. In the non-FHF cohort, this occurred in only 7% of patients.

Discussion

In this retrospective review of cases from a single, tertiary referral center, the simplified IAIHG criteria clearly demonstrate high specificity for the probable and definite diagnosis of AIH in patients with a nonfulminant course. However, despite sensitivity remaining high for a probable diagnosis it diminishes to only 70% for a definite diagnosis of AIH. Additionally, the concordance between the 1999 and simplified criteria was poor for both probable and definite AIH with the 1999 criteria producing a greater area under the ROC curve, consistent with greater diagnostic accuracy. Of further concern is the lack of sensitivity evident on the simplified criteria for a definite diagnosis of AIH, when it is clearly demonstrated that virtually all patients with AIH (99.5%) in this cohort were classified as definite AIH by the 1999 criteria.

A recent, single-center study corroborates the high sensitivity of the simplified criteria for an overall diagnosis of AIH (95%) but notes it to be inferior to the 1999 criteria (100%). Specificity remained high in this retrospective series.9 In addition, this report suggests that the simplified criteria perform less well in patients with atypical features, a statement with which we concur, based on our own findings. This study, however, appears intended to compare the performance of the 1999 and simplified criteria and, furthermore, does not represent a true validation of the simplified criteria, because an unspecified number of patients in this study were included in the IAIHG simplified criteria cohort reported by Hennes et al.8 Consequently, the close similarity in sensitivity and specificity between the two reports is entirely to be expected and therefore it is reasonable to conclude that the simplified criteria have not, until now, been independently validated.

In the published iteration of the simplified IAIHG criteria, the authors included variant or overlap syndromes as representing AIH for the purposes of their study, because it was deemed important to detect the AIH component in these syndromes. This, however, is a moot point, and open to debate, since variant syndromes may behave differently over time from classical AIH.15 In keeping with the simplified IAIHG group report guidelines,8 we also included variant syndromes as representing AIH, so that a true comparison could be undertaken. Although the sensitivity (100% for 1999 criteria, 93% by simplified criteria) and specificity (97% by 1999 criteria, 98% on simplified criteria) for an overall diagnosis of AIH did not change significantly when variants were excluded, this may be due to a type II error. Possible explanations for the loss of sensitivity evident upon the simplified criteria can be evaluated by comparing the different characteristics of patients with pure AIH or variant AIH. Such analysis reveals that patients with pure AIH were more likely to be icteric, have a higher bilirubin, AST, and international normalized ratio at presentation and a greater likelihood of response to corticosteroids than variant syndrome patients. In view of their different clinical behaviors, this brings into further question whether these individuals should be specifically regarded as having AIH or not for the purposes of assessment by the simplified criteria.

If variant syndromes are thus considered as not representing AIH, our experience suggests that the main discriminators between a false negative and true positive diagnosis of AIH via application of the simplified criteria are significantly lower IgG concentrations and frequencies of positive autoantibodies, both at titers of ≥1:40 and ≥1:80. That there should be such widespread discrepancies between false negative and false positive cases for two of the four simplified diagnostic criteria is a cause for concern.

Furthermore, that the simplified criteria appear less likely to ascribe a diagnosis of AIH than the 1999 criteria, in fulminant and nonfulminant presentations, must also be interpreted in the context of exclusion of other discriminating information such as sex, alcohol and drug history, the presence of cholestatic liver tests, and perhaps most importantly, response to corticosteroids. Such a response to therapy remains a characteristic feature of AIH in clinical practice and, indeed, this finding has been strongly supported in previous incarnations of the diagnostic criteria.1, 5

It is somewhat surprising then, that responsiveness to corticosteroids was omitted from the simplified criteria. Indeed, in this cohort, the application of the simplified criteria was less likely to ascribe a definite diagnosis of AIH than would the 1999 criteria. In such a situation, a trial of corticosteroid therapy would likely be undertaken. Furthermore, the putative addition of responsiveness to corticosteroids significantly increased the sensitivity of a diagnosis of AIH. Importantly, among the non-AIH cohort, no additional patient was up-scored to probable or definite AIH via this modification.

The utility of the IAIHG scoring systems for diagnosis of AIH have not been specifically studied in patients with FHF, largely due to its low incidence but also as a consequence of the difficulties in ascribing a firm diagnosis to patients with this disorder. However, data published from the United States Acute Liver Failure Study Group, reports that almost 49 of 125 patients (39%) with non–acetaminophen induced acute, or subacute liver failure, were due to either AIH (10%) or an indeterminate etiology (29%), with a proportion of these patients likely to have undiagnosed AIH.16

In support of this statement, and as previously discussed, autoantibodies may be identified in patients with FHF. In the study by Bernal et al., autoantibodies were not identified in any patient with acetaminophen-induced FHF but were present in 43% of non–acetaminophen induced etiologies.10 Moreover, the more specific soluble liver antigen was the most frequent antibody identified in this series.10 The presence of these AIH-associated autoantibodies in patients with FHF does not, however, prove causation, and indeed may be present in patients with hepatotropic viral infections.17, 18 Nonetheless, the presence of autoantibodies in patients with FHF remains of clinical interest, because many of these patients demonstrate features associated with autoimmunity such as female sex and the presence of human leukocyte antigen DR3 allotype.19, 20

Although there is indeed some overlap between our FHF cohort and that evaluated by Bernal et al., this equated to only nine cases or 13% of the total (W. Bernal, personal communication). Importantly, in our series, the simplified IAIHG criteria appear less sensitive than the “gold standard” of the 1999 criteria, in ascribing an overall (probable or definite) diagnosis of AIH (24% versus 40%). This was entirely due to 11 patients (16% of this cohort) who were coded as probable AIH via the 1999 criteria but on the simplified criteria were classified as not having AIH. Implicit in this statement, is the observation that no patient who met the simplified criteria for probable AIH was classified as not having AIH on the 1999 criteria. In this clinical context, however, a greater proportion of patients were classified as definite AIH by the simplified criteria than by the 1999 criteria, implying that the newer system has greater specificity but, overall, inferior sensitivity. In clinical practice, this could lead to the diagnosis of AIH not being considered and therefore corticosteroid therapy not being instituted. This has clinical relevance, because the early use of corticosteroids in this setting may improve outcome.21–28 The application of the simplified and 1999 criteria to patients with FHF should, however, be interpreted with caution, because these criteria were not developed with this clinical presentation in mind, and their utility in this setting is studied for the first time in this report. Nonetheless, because the simplified criteria reported by the IAIHG aims to improve the diagnosis of AIH at its initial presentation, it remains important not to ignore the small, but highly significant, proportion of patients with a fulminant course.

A further consideration for testing of these criteria relate to a unique population of individuals who have viral liver disease and overlapping features of autoimmunity, either present at disease presentation or precipitated by the use of interferon therapy. We attempted to utilize these criteria in a cohort of 17 patients (12 with chronic HCV and five with chronic HBV) with features of autoimmunity. Among this cohort, 13 had autoimmune features (autoantibodies and/or elevated IgG) prior to antiviral therapy, without features of autoimmune liver disease on biopsy with the remaining four patients developing autoimmune features after therapy. Four patients in this cohort were ascribed a diagnosis of AIH (two being interferon-induced) by clinical perception. Interestingly, the simplified criteria only identified two of these individuals as having AIH superimposed on viral hepatitis, whereas the 1999 criteria identified all four patients, thus corroborating its superiority as a diagnostic tool in challenging cases.

In conclusion, in this study we have demonstrated that the simplified IAIHG score identifies the presence of AIH with an extremely high degree of specificity but that sensitivity is lacking, particularly for a definite diagnosis of AIH. This is important because, in a prospectively obtained cohort of patients attending our institution, 99.5% were classified as definite AIH by the 1999 criteria. Additionally, among a separate cohort of patients with FHF, a smaller proportion of patients would be ascribed a diagnosis of AIH on the simplified criteria than on the 1999 criteria, or as attributed by clinical perception. Prospective evaluation of these criteria is required to corroborate these observations.

Ancillary