Indeterminate QuantiFERON-TB Gold Increases Likelihood of Inflammatory Bowel Disease Treatment Delay and Hospitalization

Abstract

Background:

QuantiFERON-TB Gold (QFTG) is a blood test used to diagnose latent tuberculosis infection (LTBI) prior to TNF-α inhibitor (anti-TNF) initiation. We sought to determine factors associated with indeterminate QFTG results in inflammatory bowel disease (IBD) patients and whether indeterminate results are associated with IBD-related morbidity.

Methods:

This nested case-control study included IBD patients who underwent QFTG testing. Cases were patients with indeterminate QFTG and controls were those with negative QFTG. The association of demographic and clinical data with indeterminate QFTG result was assessed using logistic regression. We examined the clinical impact of indeterminate QFTG results on risk of hospitalization and delay in anti-TNF initiation using inverse probability-of-treatment weighting (IPTW) regression.

Results:

We identified 411 patients with QFTG testing (320 negative, 80 indeterminate, and 11 positive results). No patient with an indeterminate result subsequently had LTBI. Systemic corticosteroid use (OR 4.4, 95% CI 2.0 – 9.6) and hospitalization at the time of QFTG (OR 3.8, 95% CI 1.9 – 7.7) were associated with indeterminate QFTG, while immunomodulator use was nearly statistically significant (OR 3.1, 95% CI 0.9 – 9.8) and anti-TNF use was not (OR 0.9, 95% CI 0.2 – 4.6). After IPTW adjustment, indeterminate QFTG was associated with a 23.1% (95% CI 8.2 – 37.9%) greater probability of delay in anti-TNF initiation beyond 30 days and an 11.9% (95% CI 0.6 – 23.1%) greater probability of hospitalization within 60 days.

Conclusions:

Systemic corticosteroid use and hospitalization were associated with an indeterminate QFTG result. Indeterminate QFTG results were associated with delayed anti-TNF initiation and subsequent hospitalization.

INTRODUCTION

Tumor necrosis factor-α inhibitors (anti-TNFs) were approved for the treatment of Crohn’s disease in 1998 and ulcerative colitis in 2005. These diseases are collectively referred to as inflammatory bowel disease (IBD). Subsequent post-marketing follow-up of anti-TNF users revealed an increased risk of latent tuberculosis reactivation_^1,2 due to impaired granuloma formation._³ The American Gastroenterological Association, American College of Gastroenterology, European Crohn’s and Colitis Organisation, and FDA-approved medication prescribing instructions recommend screening for latent tuberculosis infection (LTBI) prior to initiating anti-TNFs._^4–9 In addition, LTBI screening has been identified as a quality of care process measure._¹⁰ Rescreening is recommended periodically, but the optimal timing and frequency have not been determined._¹¹ A recent study showed that several factors may influence appropriate use of tuberculosis screening, such as being treated at an academic hospital or receiving care in an urban setting. Additionally, the study demonstrated a low rate of reactivation, with only 2 reactivations in 7210 patient-years of follow-up._¹²

Traditionally, LTBI screening has been performed using tuberculin skin testing (TST) by intradermal instillation of tuberculin antigens. Recently QuantiFERON-TB Gold (QFTG), an interferon-gamma release assay,_¹³ has partially supplanted TST testing for LTBI screening because there is no need for patient follow-up for result interpretation, it is not subject to interpretation bias, and it has fewer false positives among Bacillus-Calmette-Guerin vaccine-exposed patients._^14,15 Additionally, the American Thoracic Society, the Infectious Disease Society of America, and the Centers for Disease Control and Prevention issued guidelines recommending use of interferon-γ release assays like QFTG for diagnosis of tuberculosis in any patient for whom “testing for LTBI is warranted.”_¹⁶ QFTG results are reported as positive, negative, or indeterminate. Both TST and QFTG tests are dependent on in-vivo interferon release via delayed type IV hypersensitivity reaction to yield a positive result. Unlike TST, QFTG testing yields an indeterminate result when patients exhibit a poor response to tuberculosis mitogen secondary to T-cell dysfunction; these indeterminate results are not interpretable for LTBI. Thus, conditions that cause indeterminate QFTG are expected to cause anergy and a potential false negative with TST testing._¹⁷ For this reason, TST is not an appropriate supplement for LTBI testing in patients with an indeterminate QFTG.

Risk factors for an indeterminate QFTG result are not well understood. Previous studies assessing clinical factors associated with an increased risk of receiving an indeterminate result have had conflicting results and low statistical power._^18–23 Furthermore, while indeterminate QFTG results generate uncertainty regarding the presence of LTBI, there are currently little data on how these results may affect patient care. It is possible that indeterminate test results lead to additional testing, consultations, and delays in initiation of immunosuppressant regimens for patients with symptomatic IBD. Given the low prevalence of LTBI in the United States,_²⁴ the costs and morbidity associated with delays in care induced by further evaluation of an indeterminate QFTG result may outweigh the benefit of a confirmed negative result, especially among patients with no risk factors for LTBI.

In this study, we sought to identify the risk factors for receiving an indeterminate QFTG result by developing a multivariable explanatory model. Additionally, we hypothesized that patients who had indeterminate QFTG would be at an increased risk of short-term IBD morbidity compared to patients with negative results.

MATERIALS AND METHODS

Study design

We conducted a retrospective nested case-control study within a cohort of patients with IBD in the University of Pennsylvania Health System who were treated from 2009 to 2014. At the time of the study, the health system was composed of three inpatient hospitals with associated outpatient clinics in Philadelphia, Pennsylvania, USA, with a unified, electronic medical record and searchable data warehouse. During the study period, there was no protocol for how physicians within the health system should respond to indeterminate QFTG.

Cohort description and inclusion criteria

Individuals ≥18 years of age with an inpatient or outpatient encounter with an International Classification of Diseases, Ninth Revision, Clinical Modification diagnostic code for Crohn’s disease (555.0 – 2, 550.9) or ulcerative colitis (556.0 – 9) and a subsequent order for QFTG (Current Procedural Terminology code 86480) were identified by querying the health system’s electronic data warehouse. Using chart review, the indication for QFTG was recorded as (1) planned initiation of an anti-TNF for flare or disease progression, (2) planned change of an anti-TNF to another anti-TNF for flare or disease progression, (3) screening for LTBI in an asymptomatic patient already on an anti-TNF, or (4) screening for LTBI in a patient with risk factors for tuberculosis exposure. We included patients already receiving an anti-TNF to assess whether anti-TNF exposure is a risk factor of indeterminate QFTG result in those undergoing repeat testing either for consideration of switching to another anti-TNF or those undergoing routine re-screening. For those who did not initiate an anti-TNF after QFTG, we performed chart review to determine the clinical reason for not starting anti-TNF.

For the nested case-control study, cases were defined as patients with an indeterminate QFTG result, and controls were defined as patients with a negative QFTG result. We reviewed the medical records of all patients with an indeterminate result on initial QFTG and a 2:1 random sample of patients with IBD with a negative result on initial QFTG. Because the aim of the study was to understand the risk factors for and clinical impact of indeterminate QFTG results relative to negative results, patients with positive QFTG were excluded from the control population.

Determination of QFTG results

The results of QFTG testing, defined as positive, negative, or indeterminate, were extracted from the electronic medical record. All QFTG samples were processed in the central laboratory of the Hospital of the University of Pennsylvania, the largest hospital of the University of Pennsylvania Health System. Absolute interferon-γ release levels were only available for positive tests. Negative results indicated interferon-γ release was less than 0.35 IU/mL. Indeterminate results indicated that interferon-γ release from the patient specimen mixed with positive-control mitogen was less than 0.5 IU/mL.

Exposures of interest

Patient characteristics collected at the time of QFTG testing included age, sex, race, continuity gastroenterologist, IBD disease classification, and anatomic areas of disease involvement. Additional clinical data collected included laboratory results obtained within 30 days prior to or 5 days after QFTG (complete blood count with differential, creatinine, albumin, erythrocyte sedimentation rate, and C-reactive protein) and whether the patient was hospitalized at the time of QFTG testing. Laboratory tests were categorized as normal versus abnormal based on University of Pennsylvania Health System laboratory reference ranges.

Medication exposure at the time of QFTG was categorized as any exposure to systemic corticosteroids (prednisone or methylprednisolone, not including budesonide), immunomodulators (azathioprine, 6-mercaptopurine, or methotrexate), or anti-TNFs (infliximab, adalimumab, certolizumab pegol, or golimumab) at the time of testing, as defined via review of the electronic medical record. The cumulative duration of corticosteroid exposure and corticosteroid dose were also assessed. Patients using integrin inhibitors (natalizumab or vedolizumab) were excluded from the analysis because these medications are not currently known to be associated with tuberculosis reactivation.

Clinical impact of indeterminate QFTG

To determine the potential impact of indeterminate QFTG results on anti-TNF initiation, we assessed the following clinical outcomes: initiation of anti-TNF therapy greater than 30 days and 60 days, hospitalization within 60 days, and surgery secondary to IBD complications within 60 days from the QFTG test. For anti-TNF delay, we identified the date of anti-TNF initiation. A period of at least 30 days was selected to account for several other factors that may influence anti-TNF initiation, including insurance approval and infusion scheduling. We repeated these analyses at 60 days to assess for more significant delays in initiation. In those who did not receive an anti-TNF, we reviewed the medical record to determine the clinical reason. For hospitalization and surgery, the interval of 60 days was based on quality indicators from the National Committee on Quality Assurance Insights for Improvement._²⁵ We also measured whether an indeterminate QFTG test result was associated with an increased probability of consultation to infectious disease specialists.

Statistical analysis

Relevant patient characteristics were entered into a REDCap database_²⁶ in a de-identified manner. Statistical analyses were conducted using both R (R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria) and Stata 14 (Stata Statistical Software: Release 14, College Station, TX: StataCorp LP).

Assessing factors associated with indeterminate QFTG results

Descriptive statistics were used to compare the baseline characteristics of patients with indeterminate QFTG and negative QFTG. Univariable logistic regression was employed to assess clinical and demographic factors associated with an indeterminate result. All covariates were included in the initial multivariable model. Significant covariates were determined using sequential backward elimination of variables not associated with indeterminate testing (p >0.05). The impact of combining corticosteroids with other medications, including immunomodulators, anti-TNFs, or both immunomodulators and anti-TNFs was examined using interaction terms. Complete case analysis was used to develop the regression model. Laboratory values were not included in the multivariable logistic regression as only a subset of patients had full results available.

Because variation in QFTG sample processing has been shown to impact the rate of indeterminate results,_^27–29 we assessed for changes in the rate of indeterminate QFTG over time using χ² test for trend.

Assessing the clinical impact of indeterminate QFTG results

The association between indeterminate QFTG and clinical outcomes was initially assessed using Pearson’s χ₂ test. We then performed propensity score adjustment using inverse probability of treatment weighting (IPTW) to assess these associations while also adjusting for multiple potential confounders measured in the initial case-control study.^_30,31 IPTW is a form of propensity score analysis that balances potential confounders of an association by creating a weight based on the likelihood of having received an exposure—in this case, having received an indeterminate QFTG. As factors related to disease severity and current treatment may influence the probability of an indeterminate QFTG, we used IPTW to adjust for these individual characteristics and ensure they are equally balanced among comparisons between those with an indeterminate QFTG and a negative QFTG. By controlling for demographic, disease-related and treatment-related factors, the measured change in average treatment effect (ATE) reflects the effect of the QFTG test result, adjusted for the influence of the other confounders included in the IPTW model. The covariates used in the IPTW model were IBD subtype, age, race, hospitalization status, concurrent use of anti-TNFs, immunomodulators, combination therapy, or corticosteroids at the time of QFTG testing, and ordering gastroenterologist. Balancing of covariates before and after weighting was assessed using the pbalchk routine in Stata, supplemented with manual calculations of standardized mean differences and graphical methods where appropriate.^_32,33 Acceptable imbalance criteria of standardized mean differences were defined as <25% for standard criteria and <10% for stringent criteria.^_33,34 The relative impact of indeterminate testing on rates of subsequent delays in medication initiation, hospitalization, and surgery were then individually calculated using the teffects ipw routine in Stata.

Sensitivity analyses

Several additional sensitivity analyses were conducted. In order to assess the impact of using a stringent p-value cut-off of less than 0.05 in univariable and multivariable models of predictors of an indeterminate QFTG result, we repeated our analyses using a more inclusive value of <0.10 to determine if any additional predictors would be significant.

We also repeated our IPTW adjusted analyses in several subgroups within our primary cohort, including (a) only those who received a new anti-TNF within 365 days of QFTG testing, (b) only those with evidence of IBD flare by clinical documentation at the time of testing, (c) those who did not initiate an anti-TNF within 365 days of QFTG testing, (d) those who were not using anti-TNF therapy at the time of QFTG, (e) those not using steroids at the time of QFTG testing, and (f) those not hospitalized at the time of testing. Within the cohort of individuals receiving an anti-TNF within 365 days, IPTW-weighted Kaplan-Meier methods were employed to assess the time to anti-TNF initiation. Wald χ₂ statistics for this relationship were calculated using Cox regression.

Ethical Considerations

The study protocol was approved by the Institutional Review Board of the University of Pennsylvania.

RESULTS

We identified 411 IBD patients who had undergone QFTG testing at least once during the study period. Seventy-four patients underwent at least one repeat QFTG. Among 49 patients with repeat testing who were initially QFTG negative, there was one instance of seroconversion to positive QFTG (2%). Among 23 patients with repeat testing who were initially QFTG indeterminate, there were no instances of seroconversion to positive and fifteen instances of seroconversion to negative (Figure 1).

Figure 1.

Initial and repeat QFTG results stratified by initial test result

The median time between the initial QFTG and the first repeat was 350 days (n = 73, interquartile range (IQR) 86 – 603). The median interval for those with an initial negative QTFG result was 378 days (IQR 275 – 562), compared to 243 days (IQR 11 — 790) for those with an initial indeterminate result. The median time between the first repeat and the second repeat was 240 days (n = 10, IQR 98 – 670).

Factors associated with indeterminate QFTG

Baseline characteristics of those with indeterminate test results and randomly selected patients with negative results are presented in Table 1. In univariable analyses, a diagnosis of ulcerative colitis (OR 3.0, 95% CI 1.6 – 5.4) was associated with increased odds of indeterminate testing (Table 2). Patients with an indeterminate QFTG were more likely to be hospitalized at the time of QFTG (OR 4.6, 95% CI 2.6 – 8.4) and to use systemic corticosteroid monotherapy (OR 6.2, 95% CI 2.9 – 13.1). In univariable analysis, the duration of steroid therapy prior to QFTG testing did not impact the probability of a subsequent indeterminate result when compared to those not receiving corticosteroids. Patients using systemic corticosteroids for less than 7 days (OR 6.5, 95% CI 2.9 – 14.8), 7 – 28 days (OR 4.5, 95% CI 1.9 – 10.6), or greater than 28 days (OR 6.4, 95% CI 3.0 – 13.4) were all more likely to have an indeterminate QFTG than patients not using systemic corticosteroids. The dose of systemic corticosteroids also did not affect the probability of an indeterminate result. Both prednisone-equivalent doses <20 mg (OR 4.2, 95% CI 1.8 – 9.9) and >20 mg (OR 6.6, 95% CI 3.5 – 12.6) were associated with an increased risk of indeterminate QFTG result relative to patients not using systemic corticosteroids. The use of immunomodulators (OR 2.0, 95% CI 0.6 – 6.0) or anti-TNFs (OR 0.7, 95% CI 0.1 – 3.2) at the time of QFTG was not associated with an indeterminate result. Analysis of all medication classes in combination revealed a significant association with indeterminate QFTG only when systemic corticosteroids were among the medications (Supplemental Table 1).

Table 1.

Characteristics of Patients with Negative and Indeterminate QFTG

	Negative QFTG (n = 160)	Indeterminate QFTG (n = 80)
Gender
male	73 (46%)	38 (48%)
female	87 (54%)	42 (52%)
Race
Black	54 (34%)	15 (19%)
Unknown	4 (3%)	1 (1%)
White	90 (56%)	61 (77%)
Asian	3 (2%)	1 (1%)
Not categorized	9 (5%)	2 (2%)
IBD type
Crohn’s disease	122 (76%)	41 (51%)
SB involvement besides TI	30	14
TI involvement	46	21
Colonic involvement	67	31
Not documented	6	1
Ulcerative colitis	34 (21%)	34 (42%)
Proctitis alone	0	1
Left colon	7	2
Pancolitis	22	29
Not documented	5	2
IBD unclassified	4 (3%)	5 (7%)
Age
<45	88 (55%)	54 (67%)
45 – 65	56 (35%)	19 (24%)
>65	16 (10%)	7 (9%)
Hospitalized at QFTG
Not hospitalized	128 (80%)	37 (46%)
Hospitalized	32 (20%)	43 (54%)
Medications
No SCS, IM, or anti-TNF	67 (42%)	12 (15%)
Systemic corticosteroids	52 (33%)	59 (74%)
Immunomodulators	33 (21%)	18 (23%)
Anti-TNFs	30 (19%)	12 (15%)

NOTE SCS, systemic corticosteroids; IM, immunomodulatory; SB, small bowel; TI, terminal ileum

Table 2.

Association between baseline factors and risk of indeterminate QFTG

	OR (95% CI)
	Univariable	Multivariable
Gender
Male (ref)	1.0
Female	0.9 (0.5 – 1.6)
Race
Not categorized (ref)	1.0
White	2.7 (0.3 – 24.8)
Black	1.1 (0.1 – 10.7)
Asian	1.3 (0.1 – 31.1)
Other	0.9 (0.1 – 12.9)
IBD type
Crohn’s disease (ref)	1.0
Ulcerative colitis	3.0 (1.6 – 5.4)
IBD unclassified	3.7 (0.9 – 14.5)
Age
<45 (ref)	1.0
45 – 65	0.6 (0.3 – 1.0)
>65	0.7 (0.3 – 1.8)
Hospitalization at time of QFTG
Not hospitalized (ref)	1.0	1.0
Hospitalized	4.6 (2.6 – 8.4)	3.8 (1.9 – 7.7)
Medications
No SCS, IM, or anti-TNFs (ref)*	1.0	1.0
Systemic corticosteroids only	6.2 (2.9 – 13.1)	4.4 (2.0 – 9.6)
Immunomodulator only	2.0 (0.6 – 6.0)	3.1 (0.9 – 9.8)
Anti-TNF only	0.7 (0.1 – 3.2)	0.9 (0.2 – 4.6)
IM and anti-TNF	0.8 (0.1 – 7.1)	1.2 (0.1 – 11.3)

NOTE Ulcerative colitis was no longer significant in the multivariable model; SCS, systemic corticosteroids; IM, immunomodulator

In multivariable analysis, systemic corticosteroids (adjusted OR (aOR) 4.4, 95% CI 2.0 – 9.6) and hospitalization at the time of testing (aOR 3.8, 95% CI 1.9 – 7.7) remained significantly associated with indeterminate QFTG results (Table 2). Patients who were hospitalized and receiving systemic corticosteroids had greater odds of having an indeterminate result than either exposure alone (aOR 10.9, 95% CI 4.8–24.7) (Supplemental Table 2). When repeating our analyses including covariates with a p-value < 0.10 in the final model, the results did not differ from the primary analysis in which covariates with p < 0.05 were included.

Univariable logistic regression for the association between laboratory values revealed a statistically significant association for elevated leukocyte, neutrophil, and platelet counts with indeterminate QFTG. Anemia, low lymphocyte count, and hypoalbuminemia were also significantly associated with an indeterminate QFTG result (Supplemental Table 3).

Impact of indeterminate QFTG on clinical care

More patients with indeterminate QFTG had an interval of at least 30 days from QFTG to anti-TNF initiation compared to patients with negative results (26.3% vs 11.3%, p < 0.01). Patients with an indeterminate result were also more likely than those with a negative result to have an IBD-related hospitalization within 60 days of QFTG (23.8% vs 9.5%, p < 0.01, Table 3). After employing propensity score adjustment via IPTW to adjust for potential medication and disease-related confounders, individuals with an indeterminate QFTG were 34.7% (95% CI 20.4 – 49.1%) more likely to experience a delay of longer than 30-days before anti-TNF initiation and 11.9% (95% CI 0.6 – 23.1%) more likely to be hospitalized within 60 days. Covariate balance before and after IPTW weighting are shown in Supplemental Table 4.

Table 3. Inflammatory bowel disease clinical outcomes by QFTG result.

Clinical outcomes by QFTG result for the entire cohort and the subgroup of patients who received an anti-TNF within 365 days and selected subgroups of patients. p-value from χ2 test. Clinical outcomes are not mutually exclusive, so an individual patient may have experienced multiple outcomes. ATE: average treatment effects, Neg: Negative, Indet: Indeterminate

	Entire cohort
Full cohort	Neg (n = 160)	Indet (n = 80)	p (χ2)	ATE % (95% CI)
Anti-TNF initiation > 30 days	18 (11.3%)	21 (26.3%)	<0.01	34.7 (20.4 – 49.1)
Anti-TNF initiation > 60 days	1 (0.6%)	10 (12.5%)	<0.01	23.1 (8.2 – 37.9)
Hospitalization within 60 days	15 (9.5%)	19 (23.8%)	<0.01	11.9 (0.6 – 23.1)
IBD surgery within 60 days	7 (4.5%)	8 (10.3%)	0.09	3.9 (−4.0 – 11.8)
Infectious Diseases consult	6 (3.8%)	19 (23.7%)	<0.01	21.5 (10.8 – 32.2)
Subgroup of patients who received anti-TNF within 365 days of QFTG	Neg (n = 75)	Indet (n = 48)	p (χ2)	ATE % (95% CI)
Anti-TNF initiation > 30 days	18 (24.0%)	16 (33.3%)	0.26	36.4 (18.5 – 54.4)
Anti-TNF initiation > 60 days	1 (1.3%)	5 (10.4%)	0.02	17.9 (3.6 – 32.1)
Hospitalization within 60 days	10 (13.5%)	15 (31.3%)	0.02	10.9 (−5.2 – 27.0)
IBD surgery within 60 days	3 (4.2%)	6 (13.0%)	0.07	6.0 (−3.6 – 15.7)
Infectious Diseases consult	1 (1.3%)	10 (21.7%)	<0.01	33.7 (11.0 – 56.5)
Subgroup of patients who underwent QFTG for IBD flare or progression of IBD	Neg (n = 126)	Indet (n = 68)	p (χ2)	ATE (%) (95% CI)
Anti-TNF initiation > 30 days	15 (11.9%)	20 (29.4%)	<0.01	42.4 (28.1 – 56.7)
Anti-TNF initiation > 60 days	0 (0%)	10 (14.7%)	<0.01	29.0 (11.9 – 46.1)
Hospitalization within 60 days	12 (9.5%)	17 (25.0%)	<0.01	7.8 (−3.5 – 19.1)
IBD surgery within 60 days	7 (5.6%)	8 (11.8%)	0.12	4.2 (−5.5 – 13.8)
Infectious Diseases consult	3 (2.4%)	15 (22.1%)	<0.01	22.7 (10.7 – 34.8)
Subgroup of patients who were not using anti-TNF therapy at the time of QFTG	Neg (n = 129)	Indet (n = 68)	p (χ2)	ATE % (95% CI)
Anti-TNF initiation > 30 days	16 (12.4%)	19 (27.9%)	<0.01	36.9 (22.8 – 51.0)
Anti-TNF initiation > 60 days	1 (0.8%)	9 (13.2%)	<0.01	24.4 (10.0–38.9)
Hospitalization within 60 days	11 (8.5%)	17 (25.0%)	<0.01	10.8 (0.2 – 21.4)
IBD surgery within 60 days	6 (4.7%)	8 (11.8%)	0.07	5.6 (−3.1–14.3)
Infectious Diseases consult	5 (3.9%)	16 (24.6%)	<0.01	22.6 (11.0 – 34.3)
Subgroup of patients who were not using steroids at the time of QFTG	Neg (n = 91)	Indet (n = 21)	p (χ2)	ATE % (95% CI)
Anti-TNF initiation > 30 days	14 (15.1%)	10 (47.6%)	<0.01	48.9 (27.7 – 70.0)
Anti-TNF initiation > 60 days	1 (1.1%)	5 (23.8%)	<0.01	33.4 (13.0 – 53.7)
Hospitalization within 60 days	5 (5.4%)	5 (23.8%)	<0.01	9.7 (−11.9 – 31.4)
IBD surgery within 60 days	4 (4.3%)	4 (19.1%)	0.02	1.6 (−5.4 – 8.6)
Infectious Diseases consult	3 (3.2%)	9 (42.9%)	<0.01	24.2 (−0.2 – 48.7)
Subgroup of patients who were not hospitalized at the time of QFTG	Neg (n = 116)	Indet (n = 37)	p (χ2)	ATE % (95% CI)
Anti-TNF initiation > 30 days	13 (11.2%)	14 (37.8%)	<0.01	51.1 (31.3 – 72.8)
Anti-TNF initiation > 60 days	1 (0.9%)	5 (13.5%)	<0.01	31.7 (3.7 – 59.6)
Hospitalization within 60 days	8 (6.9%)	6 (16.2%)	0.09	3.0 (−8.1 – 14.1)
IBD surgery within 60 days	2 (1.7%)	2 (5.4%)	0.22	2.3 (−3.4 – 8.1)
Infectious Diseases consult	3 (2.6%)	8 (21.6%)	<0.01	17.0 (1.4 – 32.5)

When assessing the reason for hospitalization in the 19 patients with indeterminate QFTG who were hospitalized within 60 days of QFTG, 16 had IBD flares, 1 had C. difficile infection, 1 had a systemic infection, and 1 had elective surgery. Of the 15 patients with negative QFTG who were hospitalized with 60 days of QFTG, 8 had IBD flares, 2 had C. difficile infection, 2 had elective surgery, 1 had an infliximab infusion reaction, and 1 had a pulmonary embolism.

We then performed a subgroup analysis of the 75 (46.9%) patients with negative QFTG and 48 (60.0%) patients with indeterminate QFTG who started an anti-TNF within 365 days of QFTG. In this cohort, we again appreciated a significantly increased risk of delay >30 days and >60 days in those with an indeterminate test compared to those with a negative test. In IPTW Kaplan-Meier analysis, there was a significant difference in the time to medication initiation for patients with indeterminate QFTG relative to patients with negative QFTG (indeterminate QFTG median time to initiation:30 days, negative QFTG median time to initiation: 17 days; log-rank test: p = 0.01, Figure 2). While our effect estimate was similar for hospitalization compared to the primary analysis, it was not statistically significant (Table 3).

Figure 2.

Kaplan-Meier analysis of days to anti-TNF initiation among subgroup of patients who started an anti-TNF medication within 365 days of QFTG (log-rank test: p = 0.01). Indeterminate QFTG median time to initiation was 30 days. Negative QFTG median time to initiation was 17 days.

Of the 68 patients who underwent LTBI testing for indication of flare or progression of disease and had an indeterminate result, 26 did not start a biologic within 60 days. The reason for delay was additional outpatient work-up such as chest x-ray or TST in 7 patients, initiation of immunomodulatory monotherapy in 5 patients, patient decision not to start anti-TNF in 4 patients, improvement of symptoms in 4 patients, and systemic infection or abscess in 3 patients. Nine of 26 patients received a consultation with an infectious diseases specialist, with 3 consults occurring in the outpatient setting. The reason for delay could not be determined by retrospective review in 3 patients. Additional clinical outcomes for patients who did not start an anti-TNF within 365 days are detailed in Supplemental Table 5.

In the subgroup analysis of patients who were not on anti-TNF therapy at the time of QFTG, we appreciated increased risks of delay in anti-TNF initiation >30 and >60 days, as well as an increased risk of hospitalization using IPTW methods (Table 3).

When restricting the analysis to patients who were not on steroids at the time of QFTG, the association between indeterminate QFTG and re-hospitalization persisted in unadjusted analysis (5 of 21 patients for indeterminate, 5 of 91 patients for negative, p = 0.01). Delays in anti-TNF receipt >30 days and >60 days remained significantly associated with indeterminate QFTG (Table 3). In IPTW analyses, we appreciated a similar point estimate for increased risk of hospitalization as with our other analyses with larger cohorts, but this was not statistically significant (ATE 9.7%, 95% CI −11.9 – 31.4).

Lastly, when restricting the analysis to patients who were outpatients at the time of QFTG, the effect on delays of anti-TNF receipt at >30 and >60 days persisted (Table 3). However, the effect of indeterminate QFTG on subsequent hospitalization was not significant (6 of 37 patients for indeterminate, 8 of 116 patients for negative, p = 0.09).

Assessment of laboratory processing variation on indeterminate results

The rate of indeterminate QFTG was categorized into years and quarters to assess for potential variations in QFTG sample processing over time. No trend was appreciated by χ₂ test for trend (p = 0.16).

DISCUSSION

QuantiFERON-TB Gold (QFTG) is commonly used to evaluate LTBI among patients initiating anti-TNFs. However, there are limited data describing the risk factors and consequences of indeterminate QFTG testing. In this study, we demonstrate that systemic corticosteroid use of any duration or dose and hospitalization at the time of testing are associated with increased odds of an indeterminate QFTG result. Additionally, we demonstrate that patients who receive an indeterminate QFTG result are more likely to experience delays in initiation of their anti-TNF therapy and have a higher probability for IBD-related hospitalization within 60 days after adjusting for patient and provider factors and disease severity. To our knowledge, this is the first study to assess whether IBD-related morbidity is associated with an indeterminate QFTG result. These data support the hypothesis that uncertainty about patients’ LTBI status due to indeterminate QFTG results has a negative impact on patient care. Importantly, we also found that no person with an indeterminate test in our cohort of 80 patients was subsequently found to have latent or reactivated tuberculosis.

These data highlight an opportunity to optimize the care we provide for patients with IBD through the implementation of an improved tuberculosis screening strategy. A proactive strategy that includes performing QFTG testing at the time of IBD diagnosis, even if anti-TNF therapy is not being considered, will reduce the probability of testing when patients are more ill and require systemic corticosteroids or hospitalization. This may in turn decrease the odds of an indeterminate QFTG result. Furthermore, given the low incidence of tuberculosis in the American population,_³⁵ re-screening patients who had an negative QTFG result at the time of the IBD diagnosis and no tuberculosis risk factors, such as travel to endemic areas, imprisonment, or homelessness, may not be necessary prior to initiation of anti-TNF therapy. Such testing may actually be detrimental, inducing harm through delays in medical decision making if such a test were to be indeterminate. These strategies may serve to decrease IBD morbidity by leading to more expedient initiation of anti-TNF therapy in patients who are actively suffering from moderate-to-severe disease, while also reducing unnecessary healthcare expenditures unlikely to improve patients’ outcomes. Further research is required to determine the feasibility, safety, and effectiveness of these approaches.

This study has several important strengths. Our large sample size, particularly the number of indeterminate results, allowed us to assess clinical and medication exposures with more granularity than previous studies (Supplemental Table 6). Our sample size also facilitated analyses of clinical outcomes while adjusting for potential confounders using propensity score-based modeling methods.

As with all observational studies, there are several potential caveats to consider when interpreting these data. The observational nature of these data limits the ability to conclude causation of systemic corticosteroid use and hospitalization at the time of testing on indeterminate QFTG. However, after controlling for multiple confounders and assessing for interaction, we identified independent associations between both systemic corticosteroids and hospitalization and subsequent indeterminate QFTG. In the case of systemic steroid use, it is possible that helper T cell activity is diminished, causing an anergic reaction when the patient’s serum is exposed to the QFTG positive control mitogen. There is not a clear mechanistic explanation for the association of hospitalization with indeterminate QFTG. Increased disease severity or other co-morbid medical illnesses present in those who are hospitalized may have potentially confounded the relationship between hospitalization and indeterminate QFTG. It is possible that there may have been an unappreciated systematic difference in the way inpatient and outpatient samples were processed that influenced the effect of hospitalization on indeterminate QFTG result. Importantly, we did not detect any differences in indeterminate QFTG result rates over time to suggest a specimen-processing or assay-related effect.

We employed IPTW to adjust for multiple patient, provider, and disease-related confounders when assessing the association between an indeterminate QFTG and subsequent clinical outcomes. However, as with any propensity-score based methodology, it is possible that unmeasured variables that were not included in these models may have influenced our results. Therefore, it remains possible that disease severity may have biased our results. However, it is reassuring that the association we appreciated between indeterminate QFTG results and delays in anti-TNF initiation were statistically significant in all subgroup analyses. The association between indeterminate QFTG results and subsequent hospitalization within 60 days was statistically significant in most analyses and had similar effect size in all analyses except our subgroup analysis focused on those who were not hospitalized at the time of initial QFTG. While this may be a statistical artifact from repeated testing in smaller subgroups, it is also possible that this could be secondary to depletion of patients at high risk for IBD-related morbidity, thereby requiring a larger population to adequately rule out a smaller effect size.

Further research is necessary to determine the optimal strategy for anti-TNF initiation amongst patients who receive indeterminate QFTG. Repeating QFTG or performing TST are unlikely to provide useful results unless factors associated with anergy such as systemic corticosteroid use and systemic illness have been mitigated. Although our study did not assess the performance of T-SPOT.TB, another interferon-based assay, the results of this study may potentially be extrapolated to centers using it for LTBI screening among immunosuppressed patients with IBD, as it has similar testing parameters to QFTG._^23,36,37 Although a negative TST cannot confirm absence of LTBI in a patient with indeterminate QFTG because of possible anergy, many patients in our cohort subsequently underwent TST to complete insurance authorization for anti-TNF approval, potentially highlighting the clinician’s interpretation of a low pre-test probability of LTBI. Additionally, considering whether the patient has risk factors for LTBI, such as imprisonment, homelessness, healthcare occupation, travel to endemic areas, or suspicious chest x-ray,_³⁸ may help the clinician better determine the pre-test probability of an individual’s tuberculosis exposure risk, providing additional valuable information in deciding whether it is safe to initiate anti-TNF therapy. Lastly, further research is needed to understand how other factors in biologic initiation, such as insurance-related delays, may influence time to initiation and potentially impact IBD morbidity

In summary, we demonstrate that QFTG testing at the time of hospitalization or during systemic corticosteroid use is associated with increased odds of an indeterminate result. Our data also demonstrate that indeterminate testing is associated with delays in initiation of anti-TNFs and an increased probability of IBD-related hospitalization. Clinicians should consider earlier QFTG screening to provide the best possible care for this challenging patient population.

Abbreviations:

Anti-TNFs

Tumor necrosis factor-alpha inhibitors

ATE

Average treatment effects

IBD

Inflammatory bowel disease

IPTW

Inverse probability-of-treatment weighting

LTBI

Latent tuberculosis infection

QFTG

QuantiFERON-TB Gold

TST

Tuberculin skin testing