Impact of Blood Volume, Tube Shaking, and Incubation Time on Reproducibility of QuantiFERON-TB Gold In-Tube Assay

Abstract

Gamma interferon (IFN-γ) release assays (IGRAs) are functional assays used serially to measure the efficacy of novel tuberculosis (TB) vaccines and to screen health care workers for latent tuberculosis infection (LTBI). However, studies have shown nonreproducible IGRA results. In this study, we investigated the effects of blood volume (0.8, 1.0, and 1.2 ml), tube shaking (gentle versus vigorous), and incubation duration (16, 20, and 24 h) on the reproducibility of QuantiFERON-TB Gold In-Tube (QFT-GIT) results for 50 subjects (33 uninfected and 17 infected). The median IFN-γ TB response (TB antigen [Ag] minus nil value) was significantly higher with 0.8 ml blood (1.04 IU/ml) than with 1.0 ml (0.85 IU/ml; P = 0.002) or 1.2 ml (0.49 IU/ml; P < 0.001) for subjects with LTBI. Compared with 0.8 ml (11.8%), there were larger proportions of false-negative results with 1.0 ml (29.4%; P = 0.2) and 1.2 ml (41.2%; P = 0.05) of blood for infected subjects. Blood volume did not significantly change the proportions of positive results in uninfected controls. Compared with gentle shaking, vigorous shaking increased the median IFN-γ response in nil (0.04 versus 0.06 IU/ml; P < 0.001) and TB Ag (0.12 versus 0.24 IU/ml; P = 0.004) tubes and increased TB responses (TB Ag_vigorous minus nil_gentle) (0.02 versus 0.08 IU/ml; P = 0.004). The duration of incubation did not have a significant impact on the proportion of positive results in uninfected or infected subjects. This study identified blood volume and tube shaking as novel preanalytical sources of variability which require further standardization in order to improve the quality and reproducibility of QFT-GIT results.

INTRODUCTION

There are 2 billion individuals worldwide with latent tuberculosis (TB) infection (LTBI) (1). Treatment of LTBI is a proven strategy for preventing progression of LTBI to active disease (2, 3). Gamma interferon (IFN-γ) release assays (IGRAs) are relatively new assays developed as an alternative to the tuberculin skin test (TST) for diagnosis of LTBI. IGRAs are functional assays that measure T cell responses to Mycobacterium tuberculosis-specific antigens, such as ESAT-6, CFP-10, and TB7.7, in whole blood or blood-derived mononuclear cells (4). Two Food and Drug Administration-approved commercial IGRAs are currently available, i.e., the QuantiFERON-TB Gold In-Tube assay (QFT-GIT) (Qiagen, Carnegie, Australia) and the T-SPOT.TB assay (Oxford Immunotec, Abingdon, United Kingdom). IGRAs have improved specificity and offer logistical advantages over the TST (4). In recent years, IGRAs have replaced the TST for annual screening of health care workers (HCWs) in many occupational health programs in North America (5, 6). In addition, IGRA conversion rate is now being used as a measure of vaccine efficacy in TB vaccine trials (7).

Despite much optimism regarding commercial IGRAs, studies in the past decade have revealed significant variability in IGRA sensitivity and reproducibility (4–6, 8–10). The sensitivity of IGRAs in culture-positive active TB cases has ranged from 65% to 100% (4, 8, 9). In contact investigation studies, the sensitivity of IGRAs for LTBI in cases that progressed to active TB ranged from 40% to 100% (10). Furthermore, studies conducted with HCWs in low-incidence settings have shown highly variable IGRA results with serial testing. The rates of conversion (negative-to-positive result) using the manufacturer-recommended cutoff values ranged from 2% to 15%, and the rates of reversion (positive-to-negative result) ranged from 20% to 40% (6, 11). The conversion rates in those studies were higher than expected based on the risk of TB infection in low-incidence settings and TST conversion rates. Recent studies have identified several sources of variability related to assay manufacturing, preanalytical processing, analytical testing, and immunological boosting and modulation (12–16). However, causes of the broad IGRA variability in sensitivity and reproducibility are not fully understood.

The suboptimal reproducibility and accuracy of IGRAs have important implications for occupational health and vaccine trials performing serial testing. In HCWs, false-negative results can lead to nosocomial outbreaks and false-positive results can result in excessive treatment and additional testing (17). In vaccine trials, false results can exaggerate or underestimate the protective efficacy of a vaccine candidate. Therefore, it is critical to identify and eliminate or standardize the predictable sources of variability. In this study, we investigated the effects of blood volume, the extent of tube shaking, and the duration of incubation of tubes on TB responses with the QFT-GIT assay in a cohort of healthy HCWs, some with and some without LTBI.

MATERIALS AND METHODS

Study design.

In a prospective study recruiting HCWs, we compared quantitative and qualitative QFT-GIT results after varying (i) blood volume, (ii) tube shaking, and (iii) incubation duration. A schematic overview of the study design is shown in Fig. 1. This study was approved by the Stanford University institutional review board.

Fig 1.

Schematic overview of the study design. Subjects underwent QuantiFERON-TB Gold In-Tube (QFT-GIT) testing to investigate the effects of blood volume variability, shaking variability, and incubation duration variability on results. Ag, antigen.

Study subjects.

Healthy HCWs were recruited from the Stanford University Medical Center clinical laboratories. After consent was obtained, subjects were asked to complete a standardized self-questionnaire with information on age, gender, ethnicity, country of birth, countries visited for at least 6 months, Mycobacterium bovis BCG vaccination history, prior TST and QFT-GIT results, known exposure to M. tuberculosis, chest X-ray results, and immune status. Prior QFT-GIT results were confirmed in the laboratory information system. Volunteers were invited to enroll if they met criteria for (i) having LTBI (history of positive TST [≥10 mm] or QFT-GIT results and ≥1 TB risk factor) or (ii) being an uninfected control (born in the United States, with prior negative TST or QFT-GIT results and no TB risk factors). Volunteers were excluded if they reported signs or symptoms of active TB. The demographic data and prior test results are presented in Table 1.

Table 1.

Demographic data and LTBI risk factors for the study subjects

Parametera	No. (%) of subjects
Sex
Male	19 (38)
Female	31 (62)
Race and ethnicity
White	22 (44)
Asian	18 (36)
African/African American	1 (2)
Hawaiian/Pacific Islander	3 (6)
Indian subcontinent	5 (10)
Native American	1 (2)
Place of birth
USA	19 (38)
Foreign country	31 (62)
Previous TST result
Negative	29 (58)
Positive	17 (34)
NA	4 (8)
QFT-GIT result
Negative	29 (58)
Positive	17 (34)
NA	4 (8)

^aNA, not available. The mean age (± standard deviation) was 44 ± 13.7 years.

QuantiFERON-TB Gold In-Tube testing.

Except for the experimental modifications described below, all tests were performed according to the QFT-GIT package insert (18). Blood was collected in a 10-ml Kendall Monoject tube with a green stopper by a trained phlebotomist, and 1 ml was immediately transferred to each QFT-GIT tube. Tubes were shaken gently by inverting the tubes 10 times according to the package insert and were immediately incubated for 24 h. Plasma was separated and stored at room temperature for up to 8 h before a quantitative enzyme-linked immunosorbent assay (ELISA) for IFN-γ (measured in IU/ml) was performed on a DSX automated system (Dynex Technologies, Chantilly, VA). One IFN-γ ELISA plate was designated for all testing for each subject. Interpretation of results was done by the software provided by the QFT-GIT manufacturer. According to the manufacturer, a positive result was defined as a TB response (TB antigen [Ag] minus nil) value of ≥0.35 IU/ml and ≥25% of the nil value.

Blood volume.

One milliliter of blood was transferred into each of the QFT-GIT blood collection tubes. In addition, 0.8 and 1.2 ml of blood was transferred into two additional TB Ag tubes. The IFN-γ TB response was determined for each blood volume using the same nil result. The volume of blood drawn directly into QFT-GIT tubes also was determined. For a randomly selected group of volunteers (n = 30), blood was drawn directly into nil and TB Ag Vacutainer tubes of the same lot number at sea level until the tube appeared filled close to the indicator line, according to the package insert. Immediately thereafter, the blood volume in each tube was measured using a calibrated pipette.

Tube shaking.

One milliliter of blood was transferred into each of the two sets of nil and TB Ag tubes. One set was mixed gently by inverting the tubes 10 times according to the package insert, ensuring that the entire inner surface of the tube was coated with the blood. The second set was held in one hand and mixed by vigorously shaking the tubes with an up-and-down motion for 10 s. The blood transfer and tube shaking were performed by the same technologist for all study subjects. Immediate incubation of tubes and ELISA were performed as described above.

Incubation period.

One milliliter of blood was transferred into three sets of QFT-GIT tubes. After mixing of the contents according to the package insert, the sets were incubated at 37°C for either 16, 20, or 24 h. ELISA was performed as described above.

Statistical analysis.

A nonparametric test, the Wilcoxon signed-rank test of medians, was used to compare differences between paired results. The chi-square test was used to determine the statistical significance of differences between proportions. All statistical tests were computed for a two-sided type I error rate of 5%. Statistical analyses were performed using Prism software (GraphPad, San Diego, CA).

RESULTS

Effect of blood volume on TB responses.

According to the package insert, the standard QFT-GIT Vacutainer tubes are calibrated to draw between 0.8 ml and 1.2 ml blood at altitudes from sea level to 2,650 feet. To determine the distribution of blood volumes drawn into QFT-GIT tubes, blood volumes drawn into the nil tube and the TB Ag tube from 30 subjects were measured. As shown in Fig. 2A, the blood volumes in the nil and TB Ag tubes ranged from 0.785 ml to 1.0 ml (median, 0.923 ml; coefficient of variation [CV], 5.8%) and from 0.810 ml to 1.010 ml (median, 0.930 ml; CV, 4.7%), respectively.

Fig 2.

Effect of blood volume variability on QuantiFERON-TB Gold In-Tube (QFT-GIT) results. (A) Distribution of blood volumes drawn into nil and TB Ag tubes for 30 subjects. The horizontal line represents the median volume. Ag, antigen; CV, coefficient of variation. (B and C) IFN-γ TB Ag − nil values for 33 uninfected (B) and 17 infected (C) subjects tested with the indicated blood volumes in the TB Ag tubes. The assay cutoff value for positive results (TB Ag − nil value of ≥0.35 IU/ml) is marked with dashed lines. Values of <0 IU/ml are shown as 0 IU/ml. The Wilcoxon signed-rank test was used to compare differences in medians.

To investigate the effects of blood volume variability on QFT-GIT results, IFN-γ TB responses (TB Ag minus nil) with 0.8, 1.0, or 1.2 ml of blood per TB Ag tube and 1.0 ml per nil tube were measured. Results were available for the three blood volumes for 50 subjects (33 uninfected and 17 infected) (Table 2). In uninfected volunteers, the median TB response was significantly higher with 0.8 ml blood (0.01 IU/ml) than with 1.0 ml (0 IU/ml; P = 0.03) or 1.2 ml (0 IU/ml; P = 0.04) (Fig. 2B). One case turned positive with 1.2 ml blood, but there was no significant difference in the proportions of positive results between the three blood volumes. In infected subjects, the median TB response was significantly higher with 0.8 ml blood (1.04 IU/ml) than with 1.0 ml (0.85 IU/ml; P = 0.002) or 1.2 ml (0.49 IU/ml; P < 0.001) (Fig. 2C). A total of 88.2% of infected subjects (15/17 subjects) had positive results with 0.8 ml blood, 70.6% (12/17 subjects) with 1.0 ml blood, and 58.8% (10/17 subjects) with 1.2 ml blood (Table 2). The proportion of subjects with positive results was significantly lower with 1.2 ml than with 0.8 ml (P = 0.05).

Table 2.

QuantiFERON-TB Gold In-Tube results with variable blood volumes

Subjects	Results {IU/ml [median (range)]} at blood volumes of:						Pb
	0.8 mla		1.0 ml		1.2 ml
	IFN-γ response	% positive	IFN-γ response	% positive	IFN-γ response	% positive	0.8 vs 1.0 ml	0.8 vs 1.2 ml	1.0 vs 1.2 ml
Uninfected (n = 33)	0.01 (−0.03 to 0.13)	0	0.00 (−0.10 to 0.08)	0	0.00 (−0.15 to 0.59)	3	0.03	0.04	0.06
Infected (n = 17)	1.04 (0.26 to 13.59)	88.2	0.85 (0.06 to 8.58)	70.6	0.49 (0.02 to 8.67)	58.8	0.002	<0.001	<0.001

^aIFN-γ responses indicate TB antigen − nil values.

^bP values are for TB antigen − nil values.

Effect of tube shaking on TB responses.

To investigate the effects of blood mixing variability on QFT-GIT results, tubes inoculated with equal volumes of blood were mixed either gently or vigorously. Results were available for 40 subjects (23 uninfected and 17 infected). The median IFN-γ concentrations with gentle and vigorous shaking were 0.04 and 0.06 IU/ml in the nil tube and 0.12 and 0.24 IU/ml in the TB Ag tube, respectively (Table 3). Compared with gentle shaking, there was a significant increase in median IFN-γ levels in the nil and TB Ag tubes with vigorous shaking (P < 0.001 and P = 0.004, respectively) (Fig. 3). Although the median TB response did not change significantly between gentle (0.02 IU/ml) and vigorous (0.03 IU/ml) shaking, the median TB response increased significantly when the vigorously shaken TB Ag response was paired with the gently shaken nil response (vigorously shaken TB Ag response minus gently shaken nil response; P < 0.001) and decreased significantly when the gently shaken TB Ag response was paired with the vigorously shaken nil response (gently shaken TB Ag response minus vigorously shaken nil response; P = 0.004) (Table 3 and Fig. 3). There was no significant difference in the proportions of positive results with gentle (32.5%) versus vigorous (35.0%) shaking, but the proportions of positive results increased to 42.5% for the vigorously shaken TB Ag response minus the gently shaken nil response and decreased to 27.5% for the gently shaken TB Ag response minus the vigorously shaken nil response (Table 3).

Table 3.

QuantiFERON-TB Gold In-Tube results with variable tube shaking

TB responsea	IFN-γ level {IU/ml [median (range)]}	Pb	% positive (n = 40)	Pb
NilGen	0.04 (0.02 to 0.29)
NilVig	0.06 (0.02 to 9.16)	<0.001
TB AgGen	0.12 (0.02 to 18.28)
TB AgVig	0.24 (0.02 to 31.81)	0.004
Tb AgGen − nilGen	0.02 (−0.098 to 18.22)		32.5
Tb AgVig − nilVig	0.03 (−0.99 to 22.65)	0.35	35.0	0.81
Tb AgGen − nilVig	0.00 (−7.08 to 17.48)	<0.001	27.5	0.63
Tb AgVig − nilGen	0.08 (−0.04 to 31.60)	0.004	42.5	0.36

^aVig, vigorous; Gen, gentle.

^bCompared with gentle shaking, nil_Gen, Tb Ag_Gen, and Tb Ag_Gen − Nil_Gen.

Fig 3.

Effect of tube shaking on QuantiFERON-TB Gold In-Tube results. (A and B) IFN-γ values in nil (A) and TB Ag (B) tubes with gentle and vigorous shaking for 40 subjects (23 uninfected and 17 infected). (C and D) IFN-γ TB Ag − nil values with gentle and vigorous shaking. (C) Responses for gently shaken nil and TB Ag tubes (TBAg_Gen - Nil_Gen) versus those for vigorously shaken nil and TB Ag tubes (TBAg_Vig - Nil_Vig). (D) Responses for gently shaken nil and TB Ag tubes versus those for gently shaken nil and vigorously shaken TB Ag tubes (TBAg_Vig - Nil_Gen). The assay cutoff value for positive results (TB Ag − nil value of ≥0.35 IU/ml) is marked with dashed lines. Values of <0 IU/ml are shown as 0 IU/ml. The Wilcoxon signed-rank test was used to compare differences in medians. Ag, antigen; Gen, gentle; Vig, vigorous.

Effect of incubation duration on TB responses.

According to the package insert, the QFT-GIT tubes should be incubated between 16 and 24 h. To determine the effects of incubation duration on TB responses, QFT-GIT tubes from 50 subjects (33 uninfected and 17 infected) were incubated at 37°C for 16, 20, or 24 h. The median TB responses for uninfected subjects were 0, 0, and 0 IU/ml and those for infected subjects were 0.66, 0.96, and 0.85 IU/ml with 16-, 20-, and 24-hour incubations, respectively (Table 4 and Fig. 4). Except for 16 h versus 24 h (P = 0.04) in the uninfected group, the median TB responses were not significantly different for the three incubation times for the uninfected or infected subjects (Table 4 and Fig. 4). No reversions or conversions were observed among the infected or uninfected subjects (Fig. 4).

Table 4.

QuantiFERON-TB Gold In-Tube results with variable incubation durations

Subjects	Results {IU/ml [median (range)]} at durations of:						Pb
	16 ha		20 h		24 h
	IFN-γ response	% positive	IFN-γ response	% positive	IFN-γ response	% positive	16 vs 20 h	16 vs 24 h	20 vs 24 h
Uninfected (n = 33)	0.00 (−0.15 to 0.08)	0	0.00 (−0.11 to 0.07)	0	0.00 (−0.10 to 0.08)	0	0.53	0.04	0.28
Infected (n = 17)	0.66 (0.04 to 8.97)	70.6	0.96 (0.05 to 8.58)	70.6	0.85 (0.06 to 8.58)	70.6	0.25	0.98	0.17

^aIFN-γ responses indicate TB antigen − nil values.

^bP values are for TB antigen − nil values.

Fig 4.

Effects of incubation duration variability on QFT-GIT results. IFN-γ TB Ag − nil values for 33 uninfected (A) and 17 infected (B) subjects tested with the indicated durations of incubation are shown. The cutoff value for positive results (TB Ag − nil value of ≥0.35 IU/ml) is marked with dashed lines. Values of <0 IU/ml are shown as 0 IU/ml. The Wilcoxon signed-rank test was used to compare differences in medians.

DISCUSSION

In this study, we investigated the roles of blood volume, tube shaking, and incubation duration in the reproducibility of the QFT-GIT assay. We showed that blood volume and tube shaking represent novel preanalytical sources of variability that likely contribute to discordant IGRA results for individuals undergoing serial testing. The impact of blood volume variability on TB responses is most concerning for false-negative results for patients with TB infections. We showed that, compared to 0.8 ml, inoculation of TB Ag tubes with 1.0 and 1.2 ml of blood, volumes that are within the range reported by the manufacturer, resulted in 17.6% and 29.4% reductions, respectively, in proportions of positive results for infected subjects. Although the blood volumes correlated with TB responses in uninfected subjects, there were no false-positive results with 0.8 ml blood in uninfected subjects. Thus, in addition to improving QFT-GIT reproducibility, our findings suggest that standardizing blood volumes to 0.8 ml may increase assay sensitivity. In a recent study that included adult patients with culture-confirmed TB, it was shown that 0.3 ml blood per QFT microtube (in a modified version of QFT-GIT that requires 0.3 ml blood per tube) resulted in TB responses similar to those obtained with 1.0 ml blood added to standard QFT-GIT tubes (19). Interestingly, TB responses with 0.3 ml blood were significantly higher than those with 1.0 ml when a higher antigen concentration (3 μg/ml) was used in the microtube. Altogether, these findings suggest that a smaller blood volume (or increased antigen concentration) may improve detection of TB response in infected individuals. Further studies are needed to determine the optimum blood volume for maximizing QFT-GIT sensitivity without compromising specificity. A limitation of this study includes not including an assessment of the impact of blood volume on the IFN-γ concentrations in the nil tubes.

We also confirmed heterogeneity in the volumes of blood drawn into TB Ag tubes (range, 0.810 ml to 1.010 ml; CV, 4.7%). Factors that contribute to blood volume variability include the blood pressure of the individual at the time of the blood draw, the position of the Vacutainer tube relative to the venipuncture site, and the altitude at which the blood is being drawn. Blood pressure has not been correlated with TB responses in previous reproducibility studies, but it may be an important factor that accounts for some of the within-subject variability observed in serial testing studies. The manufacturer of QFT-GIT produces two types of blood collection tubes; one tube type is calibrated for blood collection at low altitudes (sea level to 2,650 feet), while the second type is produced for use at high altitudes (3,350 and 6,150 feet). Users outside these altitude ranges (between 2,650 and 3,350 feet or above 6,150 feet) are instructed to collect blood using a syringe and transfer 1 ml to each of the three tubes. Future studies are needed to investigate the effects of altitude on blood volume drawn into standard and high-altitude QFT-GIT tubes.

The impact of rigorous tube shaking on QFT-GIT results would depend on whether the nil and TB Ag tubes are shaken identically. We showed that vigorous shaking caused a significant increase in the TB response when the vigorously shaken TB Ag response was paired with the gently shaken nil response and a significant decrease in the TB response when the gently shaken TB Ag response was paired with the vigorously shaken nil response. These results imply that, if the nil and TB antigen tubes are shaken separately as blood is drawn sequentially into the nil and TB Ag tubes, respectively, then there is a possibility that differential shaking will result in either a false-positive or false-negative QFT-GIT result, especially when the TB response is near the assay cutoff value of 0.35 IU/ml.

In the absence of a preventative TB vaccine (7), LTBI treatment remains the only proven strategy for preventing the progression of latent infection to active disease (2). Epidemic modeling studies suggest that eradication of LTBI in high-burden countries is necessary for successful elimination of TB (20, 21). Given the critical role IGRAs serve in identifying individuals who need prophylactic treatment, it is imperative that we apply the knowledge gained from reproducibility studies to improve the sensitivity and reproducibility of IGRAs. Emphasis on the rigorous standardization of all controllable variables, two of which are highlighted in this study, is necessary to optimize the QFT-GIT assay. For example, blood volume variability can be avoided by collecting blood using a syringe and transferring equal volumes of blood into all QFT-GIT tubes. A less standardized but more practical approach may include direct blood draws into the QFT-GIT tubes with close monitoring of the blood level relative to the 1.0-ml indicator line. Variable shaking can be minimized if all three tubes are simultaneously mixed gently in one hand. Alternatively, a roller mixer instrument may prove more effective for standardization of mixing. Incubation delay is another preanalytical source of variability that has been shown to negatively affect TB responses in IGRAs (12, 22–25). Immediate incubation of QFT-GIT can be achieved with the use of a portable incubator in the field or placement of an incubator in the phlebotomy station, as reported previously (12, 26, 27). However, not all sources of variability can be eliminated through assay standardization. There is a need for a “borderline zone” to account for unpredictable variations that occur due to random sources of error such as analytical factors (13).

The contribution of each of the three preanalytical variables to IGRA variability in this study proved to be mechanistically counterintuitive. For example, one might hypothesize that greater blood volume would yield greater TB responses due to greater numbers of antigen-presenting cells (APCs) and T cells. To the contrary, we found that 0.8 ml of blood produced a significantly higher TB response than did 1.0 and 1.2 ml of blood. This phenomenon is likely related to a higher ratio of TB antigen to APCs and/or T cells. The importance of antigen availability for T cell activation has been described previously for initiation of adaptive immunity, but its effect on the stimulation of effector T cells in IGRAs has not been documented (28). Similarly, one might hypothesize that the length of time T cells are stimulated with TB antigens would correlate with IFN-γ release; therefore, longer incubations of QFT-GIT tubes would lead to greater TB responses. To the contrary, we did not find significant changes in TB responses in infected subjects when QFT-GIT tubes were incubated for 16, 20, or 24 h. Our finding in infected subjects, however, may be inconsistent with a study showing a significantly lower TB response with shorter incubation times (16 to 17 h versus 23 to 24 h) (29). The enhanced IFN-γ release with vigorous shaking of the nil and TB Ag tubes also was unexpected given that, until recently, the QFT-GIT package insert instructed users to “mix the tubes by SHAKING VIGOROUSLY for 5 seconds to ensure that the entire inner surface of the tube has been coated with the blood.” The mechanism through which vigorous shaking stimulates T cells to release IFN-γ is not known, but it may occur through direct and or indirect activation of blood cells. Shear forces are well-known stimuli for dose- and time-dependent activation of platelets (30).

In summary, we identified blood volume and tube shaking as previously unrecognized preanalytical sources of IGRA variability. With many sources of variability becoming known, studies are now needed to quantify the improvements in IGRA reproducibility and accuracy after elimination of systematic sources of variability.