The Positive Predictive Value of Lyme ELISA for the Diagnosis of Lyme Disease in Children

Abstract

Using a Lyme enzyme-linked immunosorbent assay (ELISA), we demonstrated that high ELISA index values are strongly predictive of Lyme disease. In children with clinical presentations consistent with Lyme disease, ELISA index values ≥ 3.0 had a positive predictive value of 99.4% (95% confidence interval 98.1-99.8%) for Lyme disease, making a supplemental Western immunoblot potentially unnecessary.

Background

Current serologic testing for Lyme disease follows a two-tiered algorithm, an enzyme-linked immunosorbent assay (ELISA) followed by supplemental Western immunoblots if the first-tier ELISA is equivocal or positive.¹ We investigated whether quantitative Lyme ELISA results by themselves could inform clinical decision-making, without the need to wait for additional immunoblot results that may take several days. To this end, we performed a cross-sectional study of children who had a Lyme ELISA performed by a single commercial laboratory over a seven-year period and investigated the predictive ability of Lyme ELISA index values across a range of cut-points.

Methods

Study Design and Setting

We performed a retrospective cross-sectional study at a single pediatric institution in a Lyme disease endemic area. We included all Lyme ELISA tests ordered over the seven and a half year period between January 1, 2007 and June 30, 2014. The institutional review board approved the study protocol with waiver of informed consent.

Case identification

We identified all Lyme ELISA tests performed through an electronic query of the institutional data warehouse. We included all ELISA test results obtained from children ≤ 21 years of age during the study period. We included multiple tests from the same patient. We selected Lyme ELISA tests with an index value ≥ 1.0 (the cut-point for an equivocal result according to the manufacturer’s guidelines). We excluded all positive Lyme ELISAs without both IgG and IgM Western immunoblots performed from the same serum sample.

Data collection

For all included Lyme ELISA tests, we abstracted the following patient-related data from the medical record onto a structured data form: demographics (age, gender), clinical symptoms, EM rash diagnosed by the treating clinician, ELISA index value and immunoblot results. All serologic Lyme disease tests from the study institution were performed at a single commercial reference laboratory (personal communication, ARUP National Laboratories, Salt Lake, UT). This laboratory utilized the same Lyme ELISA and Lyme Western blot kits (MarDx, Trinity Biotech; Tray, Ireland) throughout the study period. As recommended by the manufacturer, the laboratory converted Lyme ELISA optical density values to “index values” by dividing the optical density value by a standardized factor. All specimens with Lyme ELISA index values ≥ 1.00 and ≤ 1.20 (equivocal) and > 1.20 (positive) were reflexively evaluated using both immunoglobulin G (IgG) and immunoglobulin M (IgM) Western immunoblots. The reference laboratory scored the immunoblots using CDC criteria.¹

Outcome measure

Our primary outcome was a case of confirmed Lyme disease defined as follows: either a physician-diagnosed erythema migrans (EM) rash or a positive two-tiered Lyme serologic test (using criteria specified by the testing laboratory) in a patient with a clinical syndrome compatible with Lyme disease. A positive two-tiered Lyme serology was defined as a Lyme ELISA index ≥ 1.0 with either a positive IgG or IgM immunoblot.¹ A negative two-tiered serology was defined as a Lyme index < 1.0 or a Lyme index ≥ 1.0 with negative IgG and IgM immunoblots. We defined the following as a clinical syndrome consistent with Lyme disease: early (single EM), early disseminated (multiple EM, cranial neuritis, meningitis, carditis) or late disseminated (large joint mono-articular arthritis).² Children with non-specific clinical presentations (e.g. fever or fatigue) were considered negative for Lyme disease, even with a positive two-tiered serology.

Statistical Analysis

We present continuous variables as medians with interquartile range and categorical variables as proportions. We selected the optimum cut-point for the Lyme ELISA index value to discriminate Lyme disease from other causes of the patient’s clinical findings and then calculated the positive predictive value (PPV).

All data were analyzed using SPSS version 21.0 (SPSS Inc., Chicago, IL).

Results

Of the 7289 Lyme ELISA tests performed, 1226 (17%) had an index value ≥ 1.0. Of these, we included the 1215 tests with supplemental Western blots performed (99% of those with a positive or equivocal Lyme ELISA). The median patient age was 10.5 years (interquartile range 7.1-14.4 years). Lyme disease tests were ordered for 87 children (7% of tests sent) with EM rash, 258 (21%) with suspected early disseminated disease, 695 (58%) with suspected late disease, and 175 (14%) with non-specific signs and symptoms.

Six hundred and eighty Lyme ELISAs (9% of the ELISAs performed) were obtained from children with Lyme disease. Of these, the diagnosis of Lyme disease was made as follows: 32 with EM rash alone, 22 with EM rash and positive two-tiered Lyme serology, and 626 with positive two-tiered serology. Of those with positive two-tiered serology, 144 (20%) were positive by IgM immunoblot only, 295 (40%) were positive by IgG immunoblot only, and 209 (40%) had both a positive IgM and IgG Western immunoblot.

Increasing the Lyme ELISA index cut-point increased the positive predictive value of the ELISA for the diagnosis of Lyme disease (Table). A Lyme ELISA index value ≥ 3.0 was strongly predictive of Lyme disease [465/468; positive predictive value (PPV) of 99.4% (95% confidence interval 98.1-99.8%)]. Of the 465 with Lyme disease, 33 (7%) had early, 77 (17%) had early disseminated and 355 (76%) had late disease. The three children with a Lyme ELISA index ≥ 3.0 and negative Western blot had knee mono-arthritis and were diagnosed with non-Lyme inflammatory arthritis.

TABLE.

Classification of patients by Lyme ELISA across a range of cut points.

Lyme ELISA Cut-point	Lyme disease	Not Lyme disease
		Negative Western Blot	Positive Western Blot*
1.00 (n=1215)	680 (56%)	499 (41%)	36 (3%)
1.21 (n=953)	651 (68%)	268 (28%)	34 (4%)
2.00 (n=635)	600 (94%)	27 (5%)	8 (1%)
3.00 (n=468)	465 (99%)	3 (1%)	0 (0%)
4.00 (n=220)	220 (100%)	0 (0%)	0 (0%)

^*With non-specific clinical presentations

Discussion

We performed a large cross-sectional study of children from a Lyme endemic region for whom two-tiered Lyme disease serologic testing was obtained. A higher Lyme ELISA index value was more strongly predictive of confirmed Lyme disease as conventionally defined. For children with a Lyme ELISA index value between 1.0 and 2.9 and a clinical presentation compatible with Lyme disease, supplemental Western immunoblot testing should still be performed. However, as nearly all children with compatible clinical syndromes and a Lyme ELISA index ≥ 3.0 had Lyme disease, clinicians may be able to make initial clinical management decisions for these children without supplemental immunoblot results.

Clinicians caring for children with possible Lyme disease often need to make initial management decisions before results from two-tiered Lyme serologic testing are available. Clinical prediction rules can assist decision-making by estimating the risk of Lyme disease in children with facial palsy,³ meningitis,^4-6 or arthritis.⁷ However, the clinical presentations of Lyme disease and its mimics can overlap considerably. A rapid and accurate first-tier diagnostic test would facilitate early initiation of appropriate therapy while limiting unnecessary invasive procedures. Antimicrobial therapy active against Borrelia burgdorferi could be promptly initiated for children with confirmed Lyme disease. Perhaps more importantly, in the appropriate clinical scenario, unnecessary invasive procedures such as arthroscopic joint wash-out for treatment of possible septic arthritis or hospital admission and parenteral antibiotic therapy could potentially be avoided for children with rapidly confirmed Lyme disease.⁸

Two-tiered Lyme disease serologic testing combines a sensitive ELISA with a specific supplemental Western immunoblot. We have demonstrated that, in a Lyme disease endemic area, a very high Lyme ELISA index value could be used to direct clinical decision-making without a confirmatory Western immunoblot. The Lyme ELISA kit utilized by our reference laboratory is a moderate complexity diagnostic test that requires approximately one hour to obtain results. Although our study institution utilizes a reference laboratory, many institutions perform a Lyme ELISA on-site allowing for more rapid access to results. An optimal Lyme disease testing strategy for endemic regions may include a real-time Lyme ELISA, followed by a confirmatory Western immunoblots for low-to-moderately positive Lyme ELISA tests.

Our study had several important limitations. First, as only Lyme ELISA tests with an index value ≥ 1.0 had confirmatory Western blots, we cannot calculate the sensitivity and specificity of the Lyme ELISA. Second, we studied a single Lyme ELISA assay with numeric results based on optical density values, which may not have a linear relationship with Borrelia burgdorferi antibody titer. Therefore, our findings may not hold true for other Lyme ELISA assays.⁹ Third, a positive IgM immunoblot may be falsely positive, particularly if the patient’s symptoms have been present for more than 1 month.¹⁰ However, when we re-classified children with a positive IgM alone with more than 1 month of symptoms as not having Lyme disease, the Lyme ELISA positive predictive value dropped only slightly (data not shown). Fifth, we conducted our study in a Lyme disease-endemic area. Children from a geographic area with a lower prevalence of Lyme disease or outside the typical season may require supplemental serologic testing even when the Lyme ELISA index value is significantly elevated.¹¹ Last, we recognize that our findings in a pediatric population may not apply to adult patients.

The burden of Lyme disease continues to increase annually. Based on a 2008 survey of seven large commercial laboratories, investigators estimate an annual incidence of Lyme disease in the hundreds of thousands each year, at a cost of 500 million U.S. dollars for diagnostic testing.¹² Appropriate clinical utilization of two-tiered Lyme serologic testing, including limiting unnecessary Western immunoblots for children from Lyme disease endemic areas, could limit unnecessary expenditures. Newer serologic assays (e.g. VlsE or C6 peptide) appear to have higher sensitivity in early Lyme disease,¹³ but have not yet been evaluated in children. Future clinical guidelines will need to provide guidance about the optimal utilization of available Lyme diagnostics.