We found unusual IFN-γ measurements in 6.2% of QFT-G and 6.4% of QFT-GIT. Technical errors are a potential source for many of the unusual IFN-γ measurements we observed. Despite the potential advantages of using whole blood IGRAs to detect Mtb
infection, these tests are more complex than the TST. While TST requires five measurements to complete one test (i.e. measurement of the volume of PPD to inject, the depth of injection, the time injected, the time delay until measuring induration, and measurement of induration size), QFT-G requires at least 136 measurements to complete one test (listed as supporting information in Table S1
). The number of measurements required to complete one QFT-GIT is reduced to 126 (listed as supporting information in Table S2
) by combining test antigens and including them in the tubes used to collect blood. With so many measurements and associated manipulations, it is easy to understand how technical errors could occur.
One scenario that could lead to VLMR for QFT-G or QFT-GIT is transposition of nil and mitogen IFN-γ measurements. Additional evidence of this possibility comes from examination of the Mitogen Response values. Of the 22 IGRAs with VLMR, 11 (50%) had a Mitogen Response <−10 IU/mL (data not shown). Similarly, VLAR could be due to transposition of measurements for nil and TB antigens. By eliminating the need to add antigens for QFT-GIT, the opportunity for transpositions is reduced, and we observed significantly less VLMR and somewhat less VLAR for QFT-GIT as compared to QFT-G. Additional evidence that some VLMR and VLAR are the result of technical errors is the rarity of their recurrence with subsequent testing. VLMR recurred in only one person and VLAR did not recur. Also, for most subjects with VLMR or VLAR, such unusual IFN-γ measurements were not seen on a simultaneously performed IGRA.
has been associated with immune suppression 
, technical errors may produce indistinguishable results. If due to immune suppression, LMR
would be expected on simultaneously performed tests, but only 13.5% of subjects who had LMR
and a simultaneously performed IGRA had it on a the simultaneously performed IGRA. Additionally, LMR
occurred in more than one test for only 21.2% of the subjects who had LMR
The same transpositions that cause LMR and VLMR could account for a portion of the unusual IFN-γ measurements classified as HNC. However, in most situations, HNC occurred where transpositions were not suspected (i.e. without LMR, VLMR or VLAR). Other technical factors may be involved as is suggested by our observations that HNC occurred less often with QFT-GIT as compared to QFT-G, that HNC was rarely seen on two simultaneously performed IGRAs for the same person, and that HNC recurrence was uncommon. These findings suggest that tests with such unusual IFN-γ measurements should be repeated.
Sensitization to mouse antigens can generate heterophile antibodies in some people. These antibodies can bind to the capture and detection antibodies used in IGRAs and generate results consistent with HNC. Elevated levels of IFN-γ are seen in various infectious diseases (e.g., tuberculosis, acquired immunodeficiency syndrome, parasite diseases), autoimmune diseases (e.g., rheumatoid arthritis, thyroiditis, systemic lupus), and in allograft rejection. HNC would be expected to occur more frequently in patients with these conditions. Day-to-day variation in disease activity or changes in the amount of heterophile antibody present may explain some difference in HNC observed with initial and repeat testing. However, day-to-day variation would not explain difference in HNC observed when two IGRAs are performed simultaneously.
Technical errors appear to contribute to IGRA variability. However, data related to IGRA variability are scarce 
. This is in part because of the complexity of these tests. For example, reproducible QFT-GIT results require accurate measurement of [IFN-γ] in 3 samples in the correct order, and the use of multiple criteria for test interpretation. While testing samples by ELISA are traditionally performed in duplicate or triplicate, the manufacturer of QFT-G and QFT-GIT recommends testing once. Our review of results of multiple tests from the same person performed simultaneously and serially provided an opportunity to recognize unusual IFN-γ measurements, most of which were aberrant and seen only in one of multiple tests examined. The observation that most unusual IFN-γ are aberrant supports the recommendation to repeat testing using a fresh sample when unusual IFN-γ are encountered 
. Additional studies comparing IGRA results performed multiple times on the same sample, performed on multiple samples collected at the same time, and performed on multiple samples collected at different times are needed to more fully assess IGRA reproducibility. Recognition of unusual IFN-γ measurements and the potential for technical errors will facilitate assessment of IGRA reproducibility.
QFT-G and QFT-GIT are occasionally interpreted as positive despite unusual IFN-γ measurements. In our study 21.4% of positive results were associated with unusual IFN-γ measurements. The observation that 51.1% of subjects with positive results associated with HNC and 61.9% of positive results associated with LMR had negative or indeterminate results by all other tests examined raises doubt as to their validity. It would seem that tests with unusual IFN-γ measurements should be interpreted as indeterminate regardless of the measured response to TB antigens.
Criteria for interpreting tests with high [Nil] have evolved. Initial criteria did not include an indeterminate category for tests with HNC 
. With FDA approval of QFT-G, indeterminate criteria were included for most tests with [Nil] over 0.7 IU/mL 
. With FDA approval of QFT-GIT, indeterminate criteria were included for tests with [Nil] over 8.0 IU/mL 
. Our observations suggest that raising the cutoff for identifying high [Nil] may result in an increased number of inaccurate interpretations.
While HNC, VLMR, and VLAR were more common with QFT-G than QFT-GIT, LMR was more common with QFT-GIT. One difference that may account for these observations is how blood is mixed with antigens. Including the antigens in the blood collection tubes reduces the complexity of QFT-GIT and reduces the opportunity for technical errors, but necessitates shaking the blood vigorously to dissolve the antigens and mix them with the blood. LMR may follow inadequate shaking due to incomplete integration of the mitogen with the blood, and excessive shaking may result in lysis of lymphocytes and reduced production of IFN-γ. Similar problems with tubes containing Mtb antigens could occur, but would be difficult to detect.
A limitation of this study is our inability to confirm the presence or absence of Mtb infection. We addressed this lack of an adequate diagnostic standard by comparing results of multiple tests from the same person performed simultaneously and serially.
In conclusion, unusual IFN-γ measurements such as HNC, LMR, VLMR, and VLAR were encountered in a small number of QFT-G and QFT-GIT, and in most cases, such measurements were not seen on simultaneously or subsequently performed tests. To avoid erroneous diagnosis of Mtb infection, QFT-G and QFT-GIT with unusual IFN-γ measurements should be repeated with another blood sample and interpreted with caution if they recur.