Hepatitis A is an acute illness caused by the hepatitis A virus (HAV). For a case of hepatitis A to be confirmed, it must meet (1) either the clinical case definition and be laboratory confirmed or (2) the clinical case definition and have an epidemiological link to a laboratory-confirmed hepatitis A case.1
Clinical criteria consist of an acute illness with discrete onset of symptoms (eg, fatigue, abdominal pain, loss of appetite, intermittent nausea, vomiting) and jaundice or elevated serum aminotransferase levels. Elevated aminotransferase level alone is not a sufficient criterion to report a positive test to the Centers for Disease Control and Prevention (CDC) in the absence of clinical signs. Laboratory criteria are based on detection of IgM antibody to HAV (IgM anti-HAV) by using enzyme-linked immunosorbent assay. IgM anti-HAV is the marker for current or recent infection and persists for shorter than 6 months after clinical symptom development,2 although some evidence indicates that IgM can persist for longer periods.3 The reported incidence of hepatitis A is highest among children aged 5 to 14 years in the United States, and no chronic infection with hepatitis A is known to exist.2
During January 2002 to December 2004, laboratories in Kentucky reported more positive IgM anti-HAV test results. Most of these patients did not have hepatitis A based on their symptoms. IgM anti-HAV positive cases with no symptoms pose a serious concern for public health officials because of potential waste of valuable resources to assess whether contacts require postexposure immunoprophylaxis. In late 2004, an investigation was initiated by the Kentucky Department for Public Health (KDPH) to determine the magnitude of the problem, characterize IgM anti-HAV positive cases with no symptoms, and evaluate risk factors for testing positive among those patients.
MATERIALS AND METHODS
Surveillance Records Review
The Kentucky Department for Public Health collects data on hepatitis A as a nationally reportable disease. We reviewed all hepatitis A surveillance records at KDPH for January 1, 2002, to December 31, 2004. For the purposes of this investigation, a false-positive case was defined as a positive IgM anti-HAV test result occurring during January 1, 2002, to December 31, 2004, in a patient with no credible evidence of actual HAV infection (ie, those with neither symptoms nor a risk factor history consistent with hepatitis A).
Laboratory Records Review
On the basis of the KDPH surveillance data review, we identified that 4 laboratories accounted for 87.8% of false-positive test results. Two different IgM anti-HAV test kits were available to commercial laboratories at the time. Of these 4 laboratories, A and B used a DiaSorin test kit; and laboratories C and D used an Abbott test kit. We obtained a list of all patients by hospital/clinic, regardless of test result, who had been tested for IgM anti-HAV during January 1, 2002, to December 31, 2004, from laboratories A to D (laboratory B provided data for January 1, 2002, to November 20, 2004) and data regarding test kit manufacturer for IgM anti-HAV detection.
We matched IgM anti-HAV-positive patients from laboratory reports with confirmed hepatitis A cases identified in the KDPH surveillance records. Those that did not match with confirmed cases in KDPH surveillance records were considered false-positive cases, unless later medical record review revealed that the patients actually experienced symptoms specific to hepatitis A.
Matched Case-Control Study
We initiated a matched case-control study to examine the differences between case patients with false-positive test results and persons testing negative for IgM anti-HAV. A false-positive case was defined the same as noted above. A control was defined as a negative IgM anti-HAV test result among those persons tested for IgM anti-HAV within the same timeframe as cases. Cases were selected from the lists provided by laboratories A to D. Three controls per case were randomly selected from the list of patients with negative IgM anti-HAV test results at the hospital/clinic where each case arose. Because each hospital or clinic used service of one specific laboratory and each laboratory used only one specific test kit at the time of diagnosis, we matched controls by hospital/clinic where each case arose by laboratory and by test kit manufacturer simultaneously. A total of 71 cases were initially selected from 10 hospitals and 15 clinics in Kentucky that use the service of 1 of the 4 laboratories. Charts of 5 case patients were not located, and these 5 patients were excluded from the study, bringing the total number of cases included in the analysis to 66. Only 193 controls were selected as a result of insufficient available records in certain facilities.
Univariate and bivariate analyses were conducted in EpiInfo (Version 3.3.2; CDC, Atlanta, GA). Statistical significance was determined by using t tests (continuous variables) and χ 2 tests (categorical variables) and was defined as a P value of <0.05. SAS (Version 9.0; SAS Institute, Cary, NC) was used to conduct conditional logistic regression analysis for case-control data. Both univariate and multivariate analyses were stratified by hospital to control for the effects of matching, and a forward selection method was used to select variables for an initial model predicting case status (false-positive test result) by using an entry criterion of P = 0.2 with the Logistic Procedure in SAS. This model was compared against other models with variables of interest, considering effects on point estimates and confidence intervals (CIs) to determine the final model of best fit. Interaction terms between variables of importance were introduced and tested for significance.
Surveillance Records Review
A total of 113 confirmed cases of hepatitis A were reported to CDC from KDPH during January 1, 2002, to December 31, 2004. A total of 156 false-positive hepatitis A cases were identified through standard reports during that same timeframe. The ratio of confirmed to false-positive cases was 1:1.4. Laboratories A to D and their respective totals of potential false-positive cases identified from surveillance records in Kentucky were as follows: laboratory A, 53 (34%); laboratory B, 48 (30.8%); laboratory C, 24 (15.4%); and laboratory D, 12 (7.7%) (as noted in Methods, we only examined data on the 4 primary laboratories testing for hepatitis A; other laboratories accounted for the additional 19 false-positive hepatitis A cases not included in the laboratory counts given above).
Laboratory Records Review
Examining test records supplied directly by the laboratories, laboratory A (using the DiaSorin test kit) tested a total of 5642 persons for IgM anti-HAV; 53 (0.94%) met the case definition for being false-positive. Laboratory B (also using DiaSorin) tested 7310 persons, yielding 70 (0.96%) false-positive cases. Laboratory C (using the Abbott test kit) tested 32,178 persons, yielding 26 (0.08%) false-positive cases. Laboratory D (also using Abbott) had the lowest rate of false-positives, with 12 (0.04%) of 29,786 persons tested. The overall false-positive rate for DiaSorin test kit was 0.95%, approximately 15 times greater than that for the Abbott test kit (0.06%). Number of tests run per year was only available for laboratories B and D; neither laboratory showed upward or downward trends over the time interval examined.
DiaSorin test kits had been used for 40 cases and 104 controls; Abbott test kits for 26 cases and 89 controls. The median age of case patients was 69 years (mean, 66 years; range, 36-91 years) versus 48 (mean, 48 years; range, 7-92 years) for control subjects (t = 7.4; P < 0.001). Twenty-five case patients (37.9%) were male versus 97 (50.3%) control subjects (adjusted odds ratio [aOR], 0.6; 95% CI, 0.3-1.1). All odds ratios are adjusted for the effects of hospital-matched controls through stratification.
The type of testing was similar between cases and controls; an acute hepatitis panel (including IgM anti-HAV, hepatitis B surface antigen, hepatitis B core antibody IgM, hepatitis C virus antibody) was the most frequent test ordered.
Reasons for testing did not differ between case and control groups. The most common reason for testing among both case patients and control subjects was past elevation of serum aminotransferase levels (Table 1).
Case patients had 4.4 (95% CI, 1.6-11.9) times the odds of having cirrhosis, 1.9 (95% CI, 0.6-5.7) times the odds of having fatty liver, and 3.0 (95% CI, 1.5-5.9) times the odds of taking medications of any kind during the week before testing for IgM anti-HAV. Case patients had 12.2 (95% CI, 3.4- 44.0) times the odds of having medical record documentation of a probable history of prior infection with HAV. Testing false-positive for IgM anti-HAV was not associated with testing positive for hepatitis B or hepatitis C virus (aOR, 0.7; 95% CI, 0.3-1.7).
Because of missing data, the final model included 57 cases and 160 controls. Only age, reported history of hepatitis A, cirrhosis, and medication use were determined to be significant in the final model, controlling for hospital through stratification (Table 2). Case patients were more likely to have a history of probable hepatitis A noted in their medical record (aOR, 9.1; 95% CI, 1.7-48.7), older than 50 years (aOR, 5.3; 95% CI, 2.1-13.6), have had a diagnoses of cirrhosis (aOR, 4.2; 95% CI, 1.2-14.7), and have taken medications of any kind during the week before testing (aOR, 3.0; 95% CI, 1.2-7.4). Because of questionable reliability of the reported history of hepatitis A variable, we also ran a model without that variable. This resulted in insignificant changes in the measures of association: older than 50 years (aOR, 5.8; 95% CI, 2.3-14.5), diagnoses of cirrhosis (aOR, 4.4; 95% CI, 1.5-12.8), and taking medications of any kind during the week before testing (aOR, 2.9; 95% CI, 1.2-6.9) each remained significant.
Multiple possibilities exist for the reported high volume of false-positive IgM anti-HAV tests: asymptomatic hepatitis A, cross-reactivity of the test with other illnesses or medications, a higher rate of false-positives for tests using the DiaSorin test kit, sheer increase in testing, an increase in the amount of inappropriate testing, or previous hepatitis A with persistent IgM. Asymptomatic illness is likely not a causative factor in these false-positive cases, barring an unusual increase in acute, asymptomatic infections (not reportable to CDC because of the requirement for both clinical signs and laboratory confirmation). False-positive cases primarily occurred among an older population where the likelihood of experiencing overt symptoms is increased for hepatitis A.4 The persons in this study did not have risk factor histories consistent with any recent likely exposures to HAV. Furthermore, asymptomatic acute infection with HAV is expected to raise alanine aminotransferase/aspartate aminotransferase (ALT/AST) levels (especially among older patients), a condition observed less frequently among case patients than among control subjects.
A higher false-positive rate for the DiaSorin test kit accounted for a large number of the false-positive IgM anti-HAV test results reported to KDPH; although the DiaSorin test kit was used in less than 20% (12,952/74,916) of the IgM anti-HAV tests analyzed, it accounted for 76% (123/161) of the false-positive test results. In the present study, assuming that all false-positive IgM anti-HAV cases were identified, the specificity of the Abbott test was 99.94%, whereas that of the DiaSorin test was 99.05%. The much higher rate of false-positives for the DiaSorin test kit highlights the fact that when tests are applied to a large population with a low rate of disease, small differences in test specificity can lead to large differences in false-positive rates. Laboratory B reported that this problem was caused by an inappropriately low cutoff value and indicated that they corrected it (EpiX report). Whether other laboratories using the DiaSorin test kit also made corrections was not investigated. If the DiaSorin test kit false-positive rate were brought down to the level of the Abbott test, a total of 117 false-positive cases would have been avoided, which would represent a 73% reduction of the total false-positive rate in the state. None of the case patients were retested by the second laboratory.
Inappropriate testing is likely also a causative factor in producing an increase in false-positive IgM anti-HAV test results. Specificities of the test kits were stated as 92.3% and 94.9% for the Abbott and DiaSorin test kits, respectively,5,6 so the specificities observed in this study (see above) were actually higher. However, when testing a population with very low rates of disease, the predictive value positive is lowered, and the proportion of false-positives increases. Ordering a hepatitis panel reduces the price of the tests, compared with ordering several individual tests, but it increases dramatically the amount of inappropriate testing.7 Screening a high number of persons for such a rare disease (113 confirmed hepatitis A cases were reported from Kentucky to CDC during 3 years and more than 60,000 tests), and especially applying the test to the population where the likelihood of disease is extremely low (older patients) increases dramatically the percentage of false-positive test results.
We detected a statistically significant and consistent association between age and testing false-positive, controlling for test kit manufacturer. This finding indicates that other factors related to age are contributing to the production of false-positive test results beyond test kit cutoff values.
Case patients were significantly more likely to be taking medication of any kind (aOR, 3.0; 95% CI, 1.2-7.4). This might lead to an increase in false-positive rates if a cross-reactivity existed between medications and the test. However, cross-reactions between medications and IgM anti-HAV tests have not been documented previously in medical literature, and no medications were associated with an increased likelihood for testing false-positive, making this explanation unlikely. Case patients are expected to have an association with higher use of medications because they are older than control subjects on average. Thus, medication use appears not to explain increased false-positive IgM anti-HAV test results.
The only medical condition associated with testing false-positive was cirrhosis, a condition that involves a chronic inflammation and fibrosis. Cross-reactivity of sera from patients with chronic inflammatory diseases has been demonstrated to produce false-positive test results for IgM against certain infectious agents, as documented in the medical literature.8 This might also occur among patients with cirrhosis who are tested for IgM anti-HAV. We were unable to determine whether cirrhosis might cause a cross-reaction leading to false-positive hepatitis A IgM test results with the current study design. We did not study the influence of rheumatoid factor in this study, but it is unlikely to cause false-positive test results: both test kit manufacturers report controlling for rheumatoid factor interference in their products.5,6
No differences were noted among case and control groups for testing reasons. Past elevation of ALT/AST was the only reason for ordering an acute hepatitis panel for most false-positive case patients and control subjects. However, ALT/AST were not elevated enough to justify ordering an acute hepatitis panel among case patients and control subjects: in a case of acute viral hepatitis, substantially higher ALT/AST levels are expected9 than those charted for these patients.
In addition, our study suggests the possibility of either persistent IgM from a previous HAV infection or asymptomatic long-term infection with HAV as a factor contributing to the increased rate of false-positives. Regarding the former, in the present study, case patients with false-positive test results were significantly more likely to have a documented history of previous hepatitis A infection in their medical record. Given that 57.8% of older persons (60-69 years old) are likely to have serologic evidence indicating previous infection with HAV,10 we cannot consider the finding of more cases having documented previous HAV infection as evidence that case patients were more likely to have been previously infected so much as having had a case that was symptomatic and noted by some patients or their clinicians. Of 10 cases with a history of hepatitis A infection, only 2 had information available regarding date of diagnosis, and both were decades before the current false-positive test. IgM might persist much longer than previously thought, possibly longer than a year.11 However, IgM persisting for decades has not been documented in the medical literature. Regarding the latter possibility, when the history of hepatitis A variable was removed from the multivariate analysis, age and cirrhosis remained the prominent risk factors. Although cirrhosis is one indicator of a chronic infection with viral hepatitis and associated with testing false-positive, this possibility cannot be addressed adequately by this study. However, hepatitis A is not thought to convert to a chronic infection. Thus, the association between cirrhosis and false-positive anti-HAV IgM test results is a consistent but puzzling finding.
This study was subject to certain limitations. We were unable to assess the actual existence of virus in the patient's serum because serum was not available. Unfortunately, clinicians who do have sera on positive IgM anti-HAV cases that are inconsistent with symptoms or risk factor history do not have a confirmatory test available with which to follow-up because polymerase chain reaction is not commercially available for analysis of HAV. One previous study demonstrated that HAV RNA was detected among 1 of 25 patients with false-positive anti-HAV IgM test results.7
We also had difficulty assessing prior history of hepatitis A because IgG status of case patients and control subjects was unknown. We relied on reports of medical history rather than IgG status. These are generally self-reports and subject to questions of reliability. Also, vaccination history against hepatitis A was not documented in the charts. However, most case patients probably had not received a vaccination, because they had been born before hepatitis A vaccination development and before the recommendations had been implemented. Further, if a positive IgM anti-HAV was the result of vaccination, either infection control or local/state health department personnel who usually conduct follow-up would have ascertained vaccination status of those patients, thus identifying a possible reason for testing positive.
Finally, we were unable to examine the effects of test kit manufacturer while controlling for other variables in the multivariate analysis because we matched controls on hospital/clinic and thus laboratory and test kit manufacturer. However, based on the great difference in test kit results, it is unlikely that controlling for other variables would have influenced the association with test kit substantially.
In conclusion, false-positive hepatitis A case patients did not appear to have acute viral infection. Therefore, public health follow-up or postexposure immunoprophylaxis of contacts is not indicated. Other factors, including an artificially high false-positivity rate in one commercial test kit and decreased positive predictive value as a result of an increase in inappropriate testing are contributors to the increased detection of false-positive IgM anti-HAV cases observed in Kentucky. To understand the association of other factors such as age and cirrhosis with testing, false-positive further research will be required.