Inflated Gamma Interferon Response with QuantiFERON-TB Gold Plus Using the Automated Liaison XL Analyzer: a Testing Algorithm To Mitigate False-Positive Results in Low-Incidence Settings

ABSTRACT

The Liaison XL chemiluminescence immunoassay (CLIA) analyzer allows total automation of gamma interferon (IFN-γ) measurement for the QuantiFERON-TB Gold Plus assay (QFT-Plus) that is used to diagnose Mycobacterium tuberculosis infection. To evaluate CLIA accuracy, plasma samples from 278 patients undergoing QFT-Plus testing were first tested with an enzyme-linked immunosorbent assay (ELISA; 150 negatives and 128 positives) and subsequently with the CLIA. Three strategies to mitigate false-positive CLIA results were investigated in 220 samples with borderline-negative ELISA results (TB1 and/or TB2, 0.1 to 0.34 IU/mL). The Bland-Altman plot of difference versus average of the two IFN-γ measurements in the Nil and antigen (TB1 and TB2) tubes showed higher IFN-γ measurements across the range of values with the CLIA than with the ELISA. Bias was 0.21 IU/mL (standard deviation, 0.61; 95% confidence interval [CI], −1.0 to 1.41). Linear regression of difference versus average had a slope of 0.08 (95% CI, 0.05 to 0.10), which was significantly nonzero (P < 0.0001). The CLIA had positive and negative percent agreement levels with the ELISA of 91.7% (121/132) and 95.2% (139/146), respectively. In borderline-negative samples tested with ELISA, CLIA was positive in 42.7% (94/220). CLIA with a standard curve resulted in 36.4% (80/220) positivity. Retesting CLIA false positives (TB1 or TB2 range, 0 to ≤1.3 IU/mL) with ELISA reduced false positives by 84.3% (59/70). Retesting with CLIA reduced the false-positive rate by 10.4% (8/77). Implementing the Liaison CLIA for QFT-Plus in low-incidence settings risks falsely elevating conversion rates and overburdening clinics and overtreating patients. Confirming borderline positives with ELISA is a viable strategy to mitigate false-positive CLIA results.

INTRODUCTION

Interferon gamma (IFN-γ) release assays, such as the QuantiFERON-TB Gold Plus (QFT-Plus; Qiagen, Hilden, Germany), are performed routinely in clinical laboratories in low-incidence settings to identify health care workers, patients, and other at-risk populations with asymptomatic Mycobacterium tuberculosis infection (1, 2). The shortage of clinical laboratory scientists combined with the drive to increase efficiency have prompted clinical laboratories in North America to embrace automation (3). Total automation of immunoassays represents an important step toward this goal. The Liaison XL is a fully automated sandwich chemiluminescence immunoassay (CLIA) with advantages over conventional semiautomated enzyme-linked immunosorbent assays (ELISAs), which require sample batching, manual sample loading, and a multipoint standard curve in each run. In 2019, the U.S. Food and Drug Administration (FDA) approved the QFT-Plus for use on the DiaSorin's automated Liaison XL platform. The Liaison CLIA would replace conventional ELISAs for quantification of IFN-γ in QFT-Plus blood collection tubes after overnight stimulation of T cells with M. tuberculosis antigens (1). Instead of a 4-point standard curve, the Liaison CLIA measures IFN-γ using a 2-point calibrator master curve, thus reducing consumable reagents and associated cost. With the exception of one study, three prior studies comparing the accuracy of Liaison CLIA to conventional ELISA systems have shown significantly higher IFN-γ values with the Liaison CLIA compared with ELISA (4–6).

The aim of this study was to evaluate the accuracy of the Liaison CLIA for measurement of IFN-γ in QFT-Plus tubes compared to a conventional semiautomated immunoassay analyzer and to investigate testing strategies that mitigate false-positive CLIA results for implementation in a low-incidence setting health system in North America.

MATERIALS AND METHODS

Ethics statement.

This study was approved by the Stanford University institutional review board. Informed consent was not necessary for discarded plasma samples.

Study design.

A total of 278 QFT-Plus tests performed from 14 September 2020 to 4 March 2022 in 176 health care workers undergoing annual tuberculosis (TB) screening at the Stanford Health Care Occupational Health Clinic and 102 patients undergoing clinical evaluation for M. tuberculosis infection at Stanford Health Care were included in this study. Plasma samples were first tested with ELISA (150 negatives and 128 positives) and subsequently tested with the Liaison CLIA. Samples were selected nonconsecutively. Negative QFT-Plus samples were selected randomly. Positive samples were selected to gather sufficient positives across the range of IFN-γ values. Qualitative and quantitative agreement levels between CLIA and ELISA were measured. After completion of the accuracy study, three testing strategies were investigated in an independent set of 220 samples with borderline-negative ELISA results (TB1 and/or TB2 minus Nil, 0.1 to 0.34 IU/mL) to determine which strategy could best mitigate false-positive CLIA results. These strategies included (i) inclusion of a 4-point standard curve with Liaison CLIA and retesting false-positive CLIA samples with (ii) CLIA and (iii) ELISA.

QFT-Plus.

QFT-Plus testing was performed in the clinical microbiology laboratory per standards of care. IFN-γ concentrations were measured on the automated robotic DSX ELISA system (Dynex Technologies, Chantilly, VA) using a 4-point standard curve and interpreted per the manufacturer’s instructions.

Liaison CLIA.

For the accuracy study, residual plasma samples (43 frozen at −70°C and 235 refrigerated) were tested on the Liaison XL platform per the manufacturer’s instructions. Refrigerated samples were tested within 72 h of ELISA testing. Frozen samples were tested months after ELISA testing.

Mitigation of false-positive CLIA results.

Borderline-negative samples tested with ELISA (TB1 and/or TB2 minus Nil, 0.1 to 0.34 IU/mL) were retested with CLIA, and three testing strategies were investigated to determine which one would best mitigate false-positive results with CLIA. First, the IFN-γ standard dilutions used to generate a 4-point standard curve with ELISA were measured with the Liaison CLIA daily, and the standard curve was used to recalculate the IFN-γ values obtained with CLIA. Second, samples with false-positive CLIA results were retested with CLIA. Third, samples with false-positive CLIA results were retested with ELISA. Retesting was done in samples with sufficient plasma volume.

Statistical analysis.

The qualitative and quantitative results obtained with the Liaison CLIA were compared to DSX ELISA results using descriptive statistics, a Bland-Altman plot, and linear regression. Statistical analyses were performed using the Prism 9 software (GraphPad, San Diego, CA).

RESULTS

Accuracy of Liaison CLIA.

The Bland-Altman plot of difference versus average of the two IFN-γ measurements per QFT-Plus tubed (Nil, TB1, and TB2) for 278 individuals showed a disagreement between the two immunoassays, where the Liaison CLIA reported higher measurements across the range of values compared with the DSX ELISA (Fig. 1A). Measured bias was overall 0.21 IU/mL (standard deviation, 0.61; 95% confidence interval [CI], −1.0 to 1.41); for the Nil tube it was 0.1 IU/mL (standard deviation, 0.31; 95% CI, −0.51 to 0.71), for TB1 it was 0.25 IU/mL (standard deviation, 0.77; 95% CI, −1.27 to 1.76), and for TB1 it was 0.27 IU/mL (standard deviation, 0.65; 95% CI, −1.00 to 1.55). Linear regression of difference versus average for all tubes had a slope of 0.08 (95% CI, 0.05 to 0.10), which was significantly nonzero (P < 0.0001). Similar results were obtained when we separately analyzed frozen and fresh plasma samples and also before and after Liaison XL instrument tuning by a manufacturer field engineer (see Fig. S1 and S2 in the supplemental material). A Bland-Altman plot of TB responses (TB1 and TB2 minus Nil) similarly showed disagreement between the two immunoassays, where the Liaison CLIA reported higher measurements of IFN-γ across the range compared with the DSX ELISA (Fig. 1B), with a bias of 0.21 IU/mL (standard deviation, 0.75; 95% CI, −1.26 to 1.68).

FIG 1.

Bland-Altman plots showing the agreement of Liaison XL CLIA with DSX ELISA for measurement of IFN-γ in the QFT-Plus assay. (A) Results for individual tubes (Nil, TB1, and TB2). (B) Results for TB1 minus Nil and TB2 minus Nil. IFN-γ measurements of >10 IU/mL in individual tubes were capped and are shown as 10 IU/mL. Average negative results are not shown.

Qualitative interpretation of results obtained with the Liaison CLIA showed positive and negative agreement levels of 91.7% (121/132) and 95.2% (139/146), respectively, with the DSX ELISA (Table 1). Details of 18 discrepant results are summarized in Table 2. Borderline positives (1, 8), defined as a TB response of 0.35 to 0.7 IU/mL in either TB1 or TB2 (1, 8), accounted for 66.7% (12/18) of discrepant results (Table 2).

TABLE 1.

Agreement and accuracy of Liaison XL CLIA with DSX ELISA for QFT-Plus samples

Liaison CLIA result	No. with indicated ELISA result			% agreement
	Positive	Negative	Total
Positive	121	11	132	91.7%
Negative	7	139	146	95.2%
Indeterminate	0	0	0
Total	128	150	278
Sensitivity and specificity (%)	94.5%	92.7%

TABLE 2.

Qualitative and quantitative results for discrepant QFT-Plus samples tested with Liaison XL CLIA and ELISA

Sample no.	Liaison CLIA result (IU/mL)								ELISA result (IU/mL)
	Result	Nil	TB1	TB2	TB1 − Nil	TB2 − Nil	MIT	MIT − Nil	Result	Nil	TB1	TB2	TB1 − Nil	TB2 − Nil	MIT	MIT − Nil
1	POS	0.17	0.34	0.89	0.17	0.72	>10.0	>10.0	NEG	0.030	0.04	0.03	0.01	0.00	>10.0	>10.0
2	POS	0.06	0.15	0.42	0.09	0.36	>10.0	>10.0	NEG	0.030	0.10	0.28	0.07	0.25	>10.0	>10.0
3	POS	0.90	1.05	2.07	0.15	1.17	>10.0	>10.0	NEG	0.030	0.02	0.03	−0.01	0.00	>10.0	>10.0
4	POS	0.11	0.46	0.07	0.36	−0.03	>10.0	>10.0	NEG	0.160	0.40	0.21	0.24	0.05	>10.0	>10.0
5	POS	0.18	0.22	0.61	0.05	0.43	>10.0	>10.0	NEG	0.110	0.13	0.31	0.02	0.20	>10.0	>10.0
6	POS	0.09	0.12	0.59	0.04	0.50	>10.0	>10.0	NEG	0.070	0.09	0.40	0.02	0.33	>10.0	>10.0
7	POS	0.11	0.08	0.50	−0.03	0.39	>10.0	>10.0	NEG	0.050	0.04	0.04	−0.01	−0.01	>10.0	>10.0
8	POS	0.04	2.43	0.06	2.39	0.02	7.34	7.30	NEG	0.040	0.05	0.04	0.01	0.00	7.16	7.12
9	POS	0.06	0.04	0.42	−0.01	0.36	>10.0	>10.0	NEG	0.020	0.02	0.02	0.00	0.00	>10.0	>10.0
10	POS	0.08	0.30	0.48	0.22	0.40	>10.0	>10.0	NEG	0.040	0.17	0.28	0.13	0.24	>10.0	>10.0
11	POS	0.02	0.05	1.82	0.03	1.80	>10.0	>10.0	NEG	0.030	0.03	0.02	0.00	−0.01	>10.0	>10.0
12	NEG	0.10	0.13	0.14	0.03	0.04	>10.0	>10.0	POS	0.04	0.05	0.59	0.01	0.55	>10.0	>10.0
13	NEG	0.05	0.12	0.13	0.07	0.08	>10.0	>10.0	POS	1.860	2.78	2.97	0.92	1.11	4.63	2.77
14	NEG	0.16	0.14	0.11	−0.02	−0.05	>10.0	>10.0	POS	0.120	0.11	0.55	−0.01	0.43	>10.0	>10.0
15	NEG	0.13	0.23	0.22	0.10	0.09	>10.0	>10.0	POS	0.090	0.48	0.17	0.39	0.08	>10.0	>10.0
16	NEG	4.66	4.52	4.65	−0.14	−0.01	>10.0	>10.0	POS	2.350	2.53	3.31	0.18	0.96	9.88	7.53
17	NEG	0.04	0.05	0.05	0.01	0.01	>10.0	>10.0	POS	0.040	0.54	0.03	0.50	−0.01	>10.0	>10.0
18	NEG	2.17	0.80	0.77	−1.37	−1.40	>10.0	>10.0	POS	0.030	0.45	0.05	0.42	0.02	>10.0	>10.0

Triplicate testing of IFN-γ standards used to generate a 4-point standard curve for DSX ELISA showed higher average values at each dilution when tested with the Liaison CLIA. A 4-IU/mL standard was measured as 4.3 IU/mL, the 1-IU/mL standard was measured as 1.31 IU/mL, the 0.25-IU/mL standard was measured as 0.37 IU/mL, and the 0-IU/mL standard was measured as 0.00 IU/mL (Table 3).

TABLE 3.

IFN-γ measurements of ELISA standards tested with Liaison XL CLIA

Concn of standard (IU/mL)	Liaison CLIA (IU/mL) in replicate
	No. 1	No. 2	No. 3	Avg	Delta
4	4.54	4.11	4.25	4.30	+0.3
1	1.32	1.27	1.34	1.31	+0.31
0.25	0.37	0.36	0.37	0.37	+0.12
0	0.00	0.00	0.00	0.00	0.00

Mitigating false-positive CLIA results.

In 220 borderline-negative samples based on ELISA (TB1 and or TB2 minus Nil, 0.1 to 0.34 IU/mL), CLIA generated 42.7% (94/220) false-positive results (Table 4). Among the false positives, the TB1 response ranged from 0 to 0.86 IU/mL (median, 0.40; interquartile range [IQR], 0.30 to 0.52) and the TB2 response ranged from 0.01 to 1.3 IU/mL (median, 0.45; IQR, 0.37 to 0.53) (Table 4). Among false-positive results, 97.8% (92/94) were in the borderline range of 0.35 to 1.0 IU/mL. Inclusion of a 4-point standard curve with Liaison CLIA resulted in a positivity rate of 36.4% (80/220) (Table 4). Retesting CLIA false positives with CLIA reduced the false-positive results by 10.4% (8/77) (Table 4). Retesting CLIA false positives with ELISA reduced the number of false positives by 84.3% (59/70) (Table 4).

TABLE 4.

Investigation of three testing strategies to mitigate false-positive Liaison XL CLIA results

Summary statistic	ELISA borderline negatives (n = 220)		Liaison CLIAa				Liaison CLIA with standard curvea				Liaison CLIA positives repeated on CLIAa		Liaison CLIA positives repeated on ELISAa
			Overall (n = 220)		Positives only (n = 94)		Overall (n = 220)		Positives only (n = 80)
	TB1 − Nil	TB2 − Nil	TB1 − Nil	TB2 − Nil	TB1 − Nil	TB2 − Nil	TB1 − Nil	TB2 − Nil	TB1 − Nil	TB2 − Nil	TB1 − Nil	TB2 − Nil	TB1 − Nil	TB2 − Nil
Positivity rate	0% (0/220)	42.7% (94/220)	NA	36.4% (80/220)	NA	89.6% (69/77)	15.7% (11/70)
Median	0.18	0.18	0.27	0.25	0.40	0.45	0.24	0.23	0.41	0.45	0.44	0.44	0.34	0.41
Range	0–0.34	0–0.34	0–0.86	0−1.30	0-0.86	0.01-1.30	0-0.96	0–1.18	0–0.96	0–1.18	0–2.98	0–1.19	0.03–-0.49	0.23–0.45
IQR	0.10–0.18	0.10–0.18	0.16–0.38	0.16–0.42	0.30–0.52	0.37–0.53	0.14–0.35	0.13–0.40	0.34–0.52	0.37–0.58	0.34–0.55	0.37–0.53	0.27–0.35	0.37–0.42

^aELISA borderline-negative samples (TB1 and/or TB2, 0.1 to 0.34 IU/mL) were tested with CLIA with and without a standard curve. False-positive CLIA samples with sufficient plasma volume were retested with CLIA and ELISA. NA, not applicable.

DISCUSSION

This study showed higher IFN-γ values in each QFT-Plus tube and higher calculated TB responses (TB1 and TB2 minus Nil) with the Liaison CLIA than with ELISA. The latter indicated that the higher Liaison Nil values did not mitigate the impact of higher antigen-induced IFN-γ values. Furthermore, qualitative interpretation of results obtained with the Liaison CLIA compared with ELISA showed a lower positive percent agreement (91.7%) compared with the negative percent agreement (95.2%). Since the majority (90 to 95%) of QFT-Plus results in our low-incidence setting are negative with ELISA, the Liaison CLIA would cause more false positives than false negatives in our low-risk population, particularly when results are borderline negative.

The findings of this study on CLIA accuracy are consistent with those of prior studies comparing the accuracy of the Liaison CLIA to conventional ELISA systems (4–6). Bisognin et al. showed significantly increased median IFN-γ values with the Liaison CLIA in TB1 (0.42 versus 0.21 IU/mL) and TB2 (0.40 versus 0.22 IU/mL) in 301 individuals compared with ELISA (4). It resulted in a sensitivity and specificity of 99.2% and 81.9%, respectively, using ELISA as the reference method (4). De Maertelaere et al. showed significantly higher TB1 and TB2 IFN-γ values with Liaison CLIA compared with ELISA in 92 individuals (6). Fernández-Huerta et al. showed significantly higher IFN-γ values with the Liaison CLIA compared with ELISA in 323 patients (5). However, in one study by Brantestig et al., a high degree of quantitative and qualitative concordance was found between the Liaison CLIA and ELISA in 125 individuals (7).

The finding that Liaison CLIA inflates IFN-γ measurement in the QFT-Plus tubes has important implications for latent TB infection (LTBI) screening in low-incidence settings. The QuantiFERON-TB Gold (QFT-Gold), the predecessor of QFT-Plus, became available in North America after it received FDA approval in 2005. Early adopters of QFT-Gold discovered falsely elevated conversion rates in low-risk health care workers and in no-risk university students compared to the conventional tuberculin skin testing (1). Consequently, some health systems had to abandon QFT-Gold and go back to skin testing (1). Since then, much has been learned about the preanalytical, analytical, and immunological factors that negatively impact the accuracy and reproducibility of QFT (1, 8). Standardization of preanalytical and analytical variables has been proposed and adopted by clinical laboratories to improve the quality of results (1, 8, 9). Adoption of a borderline range (0.2 to 0.7 IU/mL) has also been recommended to account for uncontrollable sources of variability (1, 8). The inflated IFN-γ concentrations reported by the Liaison CLIA represent a novel analytical source of variability which would falsely elevate positivity and conversion rates and trigger overutilization of health care resources and overtreatment of patients in low-incidence settings. The cause of inflated results with the Liaison XL Analyzer remains to be determined and corrected by the manufacturer, but it appears to be inherent to the detection method and not related to preanalytical factors (4, 6). Some investigators have suggested redefining the cutoff for QFT-Plus tested on the Liaison XL Analyzer (4, 5). However, it is unlikely that this approach is the correct fix given that the Liaison CLIA is falsely overestimating IFN-γ compared with ELISA values derived using an international standard. For the same reasoning, it is also incorrect to assume that the Liaison CLIA is clinically more sensitive for diagnosis of M. tuberculosis infection (5) or that it is beneficial for obtaining interpretable negative results in immunocompromised patients with indeterminate results due to a low T-cell count in their peripheral blood (4, 5). Instead, it is imperative that the manufacturer correct the overestimation of IFN-γ and prevent reporting of false positives in low-incidence settings. It is also essential that the manufacturer and/or the FDA warn clinicians and public health officials about this problem until the cause of the discrepancies is resolved.

Clinical laboratories that are interested in using the Liaison CLIA in low-incidence settings must adopt testing algorithms to mitigate false-positive results with CLIA. We showed that nearly all false-positive CLIA results in a set of samples with borderline-negative ELISA results had a TB response in the borderline-positive range of 0.35 to 1.0 IU/mL. Of three testing strategies to mitigate false-positive CLIA results, we found only repeating the false-positive samples with ELISA was effective in identifying false-positive results. Repeating with CLIA or running a 4-point standard curve with CLIA was unable to identify the false-positive CLIA results. The latter suggests that the use of a 2-point calibrator master curve instead of a 4-point standard curve is not the reason why Liaison CLIA overestimates IFN-γ measurements. Our findings are consistent with a recent study that identified a TB1 response of 0.35 to 1.55 IU/mL and TB2 response of 0.35 to 1.45 IU/mL as the range for nonreproducible positive CLIA results (10). However, our findings do not support the conclusion that retesting with CLIA can mitigate false-positive results with CLIA (10). Our findings suggest that confirmation of borderline-positive CLIA results with ELISA is a viable strategy that laboratories could use to mitigate false-positive CLIA results in low-incidence settings. The financial costs of retesting positive samples with semiautomated ELISA and having to maintain an ELISA instrument must be outweighed by the financial savings associated with using the fully automated Liaison CLIA. This strategy may be cost-effective in a low-incidence setting where <10% of samples require retesting with ELISA. However, this strategy would not be cost-effective in settings with a higher incidence of LTBI.

In summary, consistent with prior studies, we showed that the Liaison CLIA overestimates the IFN-γ measurement in QFT-Plus tubes. Clinical laboratories in low-incidence settings should be aware of this problem and avoid falsely elevating their QFT-Plus positivity rates. Confirming borderline-positive CLIA results with ELISA is a viable testing strategy to mitigate false-positive CLIA results.