Systematic review with meta-analysis: Accuracy of interferon-gamma releasing assay and anti-Saccharomyces cerevisiae antibody in differentiating intestinal tuberculosis from Crohn's disease in Asians

Background

Crohn's disease (CD) and intestinal tuberculosis (ITB) are chronic granulomatous disorders with very similar clinical, pathologic, radiologic, and endoscopic findings.1 The differentiation between these two entities can be difficult but at the same time important as misdiagnosis can lead to unwanted consequences. The initiation of immunosuppressants used in the management of CD would result in increased morbidity and mortality in patients with ITB. In contrast, the use of anti-TB therapy in patients with CD may lead to delayed diagnosis and antibiotic resistance. With the emergence of CD globally and in areas of high-TB endemicity, distinguishing CD from ITB has become a clinical challenge even in tertiary care settings and experienced hands.2, 3

Because the symptoms and signs of ITB are often nonspecific and may resemble that of CD, misdiagnosis rates between ITB and CD range from 50% to 70%.1, 4 Acid fast bacilli (AFB) on histology, smear, and culture of intestinal biopsy, which are the absolute criteria to differentiate ITB from CD, are found in only a small proportion of patients. The most reliable differentiation is to find evidence of mycobacteria TB in the intestinal tissue. However, AFB staining lacks sensitivity and specificity and culture is time consuming. Positive TB cultures are present in less than 50% of patients with ITB.4 TB polymerase chain reaction has a high sensitivity but showed variable specificity for diagnosing ITB.5, 6 In most TB-prevalent countries, a large number of ITB cases are diagnosed based on assessment of response to anti-TB therapy.

In recent years, an interferon-gamma release assay (IGRA) that measures the release of interferon after stimulation in vitro by Mycobacterium tuberculosis antigens has been approved by the Food and Drug Administration as an aid in diagnosing mycobacterial tuberculosis infection, including both latent TB and active TB disease.7 Two IGRAs, QuantiFERON-TB Gold (QFT) (Cellestis Limited, Carnegie, Victoria, Australia) and T-SPOT.TB (TSPOT) (Oxford Immunotec Limited, Abingdon, UK) are now commercially available, and their use is expanding. Among all serological markers available for inflammatory bowel disease (IBD), anti-Saccharomyces cerevisiae antibody (ASCA) has been reported to have the best combined sensitivity and specificity for CD.8

Although IGRA have been mainly used to identify latent TB infections, an increasing number of studies have also evaluated the diagnostic yield of IGRA in the differential diagnosis of ITB and CD. However, these studies have produced conflicting results. In the current study, we conducted a systematic review and meta-analysis to assess the accuracy of IGRA and ASCA in the differentiation between CD and ITB.

Methods

Results

Literature search

Among the 45 abstracts identified from the initial search, 24 met our inclusion criteria (Fig. 1). Of these 24 studies, four did not use ASCA or IGRA as a diagnostic test, four were case reports, four were duplications, and one had insufficient outcome data for analysis. Therefore, a total of 11 studies comprising 1081 participants including five on IGRA, three on ASCA, and three on both IGRA and ASCA published between 2007 and 2013 were available for the final analysis. The sample size ranged from 53 to 191 subjects. All studies were cross-sectional. Three studies were presented in abstract form, and the remaining studies were available as full text. All studies were in English language except one that was published in Chinese language. All studies were from countries in Asia (Table 1). Among the eight IGRA studies, 415 patients with CD and 290 patients with ITB were included. The mean age was 35.8 years, and 41.7% were male. The duration of follow-up ranged from six months to four years. Three studies used QuantiFERON-TB Gold, and five studies used T-SPOT.TB. Among the ASCA studies, 345 patients with CD and 258 patients with ITB were included. The mean age was 36.5, and 54.8% were male. The duration of follow-up ranged from six months to eight years.

Table 1. Characteristics of included studies

Authors/year	Site	Paper	Follow-up	IGRA used	ASCA type	No. of subjects			Mean age	No. of male	IGRA on ITB				ASCA on CD
						Overall	ITB	CD			TP	FN	FP	TN	TP	FN	FP	TN
Baek 201311	Korea	Abstract	2007–2011	T-SPOT.TB	—	95	31	59	n/a	n/a	31	0	11	48	—	—	—	—
Dutta 201112	India	Full	2006–2007	—	IgG	60	30	30	34.5	37	—	—	—	—	9	21	3	27
Ghoshal 200713	India	Full	2001–2003	—	IgG	80	16	16	38.8	46	—	—	—	—	10	6	8	8
Kim 201114	Korea	Full	2007–2008	QuantiFERON	IgG/IgA	147	75	72	38.0	86	50	25	7	65	32	40	10	65
Kim 201315	Korea	Abstract	2007–2011	QuantiFERON	—	75	22	11	n/a	n/a	13	9	0	11	—	—	—	—
Lee 201016	Korea	Full	2007–2009	T-SPOT.TB	—	60	12	44	27.3	37	12	0	6	38	—	—	—	—
Lei 201317	China	Full	2003–2011	T-SPOT.TB	IgG	191	88	103	36.6	105	36	6	5	62	5	60	0	51
Li 201218	China	Full	6 months	T-SPOT.TB	IgG/IgA	84	19	65	34.2	54	16	3	16	49	14	51	3	16
Makharia 200719	India	Full	—	—	IgG	135	30	59	35.4	54	—	—	—	—	30	29	14	16
Moon 200920	Korea	Abstract	2006–2008	QuantiFERON	—	101	30	41	n/a	29	27	3	6	35	—	—	—	—
Qui 201221	China	Full	2008–2012	T-SPOT.TB	—	53	13	20	38.9	25	12	1	5	15	—	—	—	—

• CD, Crohn's disease; FN, false negative; FP, false positive; Ig, immunoglobulin; IGRA, interferon-gamma release assay; ITB, intestinal tuberculosis; n/a, not available in the original manuscripts; TN, total number; TP, total positive.

Sensitivity and specificity of IGRA for the diagnosis of TB

Table 2 showed the pooled sensitivity and specificity estimates of eight studies of IGRA. The pooled sensitivity and specificity of IGRA for the diagnosis of ITB were 81% (95% CI: 75% to 86%) and 85% (95% CI: 81.2% to 88.6%), respectively. The SROC curve showed a good accuracy with an AUC of 0.92 (Fig. 2a). The positive likelihood ratio of IGRA was 6.02 (95% CI: 4.62 to 7.83), and the negative likelihood ratio was 0.19 (95% CI: 0.10 to 0.36); the positive predictive value of IGRA was 78% (95% CI: 72% to 83%), and the negative predictive value was 87% (95% CI: 84% to 91%). Two types of IGRA were included in this meta-analysis. QuantiFERON-TB Gold measures T cell Interferon (IFN)-gamma production in blood in response to a cocktail of early secretory antigenic target-6 (ESAT-6) and culture filtrate protein-10 (CFP-10) and TB 7.7, whereas T-SPOT.TB measures mononuclear cells from blood and the number of IFN-gamma producing cells responding to antigens including ESAT-6 and CFP-10. In subgroup analysis, the pooled sensitivity and specificity of QuantiFERON-TB Gold was 71% (95% CI: 62% to 79%) and 89.5% (95% CI: 83% to 94%), respectively. The AUC of SROC curve was 0.91. The pooled sensitivity and specificity of T-SPOT.TB was 92% (95% CI: 85% to 96%) and 83% (95% CI: 78% to 88%), respectively. The AUC of SROC curve was 0.93.

Table 2. Sensitivity, specificity, positive and negative predictive value, and positive and negative likelihood ratios of interferon-gamma release assay

Study	Sensitivity (95% CI)	Specificity (95% CI)	Positive predictive value (95% CI)	Negative predictive value (95% CI)	Positive likelihood ratio (95% CI)	Negative likelihood ratio (95% CI)
Baek 201311	100% (89–100%)	81% (69–90%)	74% (58–86%)	100% (93–100%)	5.14 (3.05–8.65)	0.02 (0.001–0.30)
Kim 201114	67% (55–77%)	90% (81–96%)	88% (76–95%)	72% (62–81%)	6.86 (3.33–14.11)	0.37 (0.27–0.51)
Kim 201315	59% (36–79%)	100% (72–100%)	100% (75–100%)	55% (32–77%)	14.09 (0.91–217.08)	0.43 (0.26–0.71)
Lee 201016	100% (73–100%)	86% (73–95%)	67% (41–87%)	100% (91–100%)	6.66 (3.24–13.67)	0.05 (0.003–0.68)
Lei 201317	86% (72–95%)	93% (83–98%)	88% (74–96%)	91% (82–97%)	11.49 (4.90–26.93)	0.15 (0.07–0.33)
Li 201218	84% (60–97%)	75% (63–85%)	50% (32–68%)	94% (84–99%)	3.42 (2.14–5.46)	0.21 (0.07–0.60)
Moon 200920	90% (74–98%)	85% (71–94%)	82% (65–93%)	92% (79–98%)	6.15 (2.91–13.01)	0.12 (0.04–0.35)
Qui 201221	92% (64–100%)	75% (51–91%)	71% (44–90%)	94% (70–100%)	3.69 (1.70–8.02)	0.10 (0.02–0.69)
Summary measures	81% (75–86%)	85% (81–89%)	78% (72–83%)	87% (84–91%)	6.02 (4.62–7.83)	0.19 (0.10–0.36)

Discussion

This meta-analysis includes the results of 11 studies and synthesizes the literature on the accuracy of IGRA and ASCA in distinguishing between ITB and CD. Our data confirmed that IGRA have a good sensitivity and specificity for the diagnosis of ITB. In particular, we found that specificity estimates for IGRA were consistently high across studies. The mean sensitivity of IGRA was 81% and specificity was 85% for the diagnosis of ITB with an AUC of 0.92. These results indicate that the overall accuracy of IGRA was relatively high for the diagnosis of ITB. We reported the positive and negative likelihood ratios which represent measures of diagnostic accuracy, and these parameters have been considered to be more meaningful in clinical practice.22 A positive likelihood ratio of 6.02 suggests that IGRA is six times more likely to be tested positive in patients with ITB than in patients without ITB. With a negative likelihood ratio of 0.19, the results suggest that patients with ITB have around 19% chance of being IGRA negative than those without ITB. The overall results remain robust when only studies with low risk of bias are included. Furthermore, the pooled results showed that the negative predictive for IGRAs was 87%, indicating a false negative rate of 13%. The relatively high negative predictive value suggests that IGRA would be acceptable for clinical purpose.

To date, IGRA have been mostly used to identify latent TB and been shown to be superior to tuberculin skin tests. The advantages of QFT-G include the lack of cross-reaction with BCG and most non-tuberculous mycobacteria, avoidance of reader bias, and the need for only a single patient visit. There are currently limited data on the role of IGRA in monitoring therapeutic responses. One study used QFT-G to follow the clinical response to anti-TB therapy in subjects with active pulmonary tuberculosis and showed a higher likelihood of converting to a negative sputum culture with decreasing titers.23

Two recent meta-analyses have reported that IGRA have high specificity for the diagnosis of latent TB especially in populations that have received BCG vaccinations.24, 25 However, currently the sensitivity of both tuberculin skin test and IGRA is suboptimal, and none of these tests could distinguish between latent tuberculosis and active disease.

The data on the usefulness of ASCA to differentiate between CD and ITB have produced more contradictory results.13 ASCA is a nonspecific antibody resulting from macromolecular transport of food antigens partly resulting from increase in intestinal permeability. Hence, any condition that increases macromolecular transport of food antigens across the intestinal mucosa may result in positive result to this test. Several authors have suggested that ASCA is not a nonspecific antibody in CD given that it is often positive in family members of patients with CD. Inferior performance of this test to differentiate between ITB and CD lies on the high frequency of positive test result even in patients with ITB.19 In our meta-analysis, six studies have reported the use of ASCA in the diagnosis of CD from ITB. Overall, the studies have shown low sensitivities and higher specificities. Given that the specificity of ASCA is higher than the sensitivity, it is more useful for the differentiation of CD from ITB than for population screening.

Three studies have reported both IGRA and ASCA in the diagnosis of ITB or CD, but only one study has analyzed the combined diagnostic validity.14 When ASCA was positive and IGRA negative, the sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio for the diagnosis of CD were 41.6% (95% CI: 30–54%), 97.3% (95% CI: 90–99%), 15.63 (95% CI: 3.87–63.0), and 0.59 (95% CI: 0.49–0.72), respectively. When ASCA was negative and IGRA positive, the combined sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio for the diagnosis of ITB were 53.3% (95% CI: 42–65%), 91.7% (95% CI: 82–96%), 6.4 (95% CI: 2.89–14.17), and 0.51 (95% CI: 0.39–0.65), respectively. However, more studies that assessed the combined values are needed before we can suggest that a combination of IGRA and ASCA can also be useful.

This meta-analysis has several limitations. First, most studies were small with heterogeneous populations. Some patients were diagnosed with ITB based on the clinical course or response to anti-TB therapy and were not confirmed by microbiological criteria. This is because we lack a gold standard test for some patients with ITB. Hence accuracy of diagnosis could lead to misclassification. Second, more than 50% of the primary studies included patients with diseases outside of ITB and CD. That is, the reported diagnostic accuracy of IGRA and ASCA was mainly for diagnosing ITB from non-ITB, and CD from non-CD. Hence, the pooled diagnostic accuracy data may not apply to the subgroup of patients who have excluded non-ITB non-CD diseases. Third, all studies were from countries with a high rate of TB incidence and may be associated with publication bias. In addition, only nine and six studies were available for IGRA and ASCA, respectively, and therefore, funnel plots and the Egger test were not performed. Future studies should involve multiple centers from different geography. Fourth, the results of ASCA may differ based on different kits and manufacturers. The positive criteria for ASCA also varies between studies depending on the type of kit used. In most studies, a positive control, negative control, and “cut-off” were run with each group of samples. A standard curve was plotted, and the optimal density of “cut-off” control was used as a marker to differentiate between positive and negative results. Fifth, although sensitivity and specificity are useful and easily measured test characteristics, they have limitations. The trade-offs between these tests characteristics are not easy to interpret. Although the pooled T-SPOT.TB sensitivity and specificity was higher than that of the QuantiFERON-TB Gold. This finding should be carefully interpreted because it is not based on direct head-to-head comparison studies. Lastly, pooled data using the combination of IGRA and ASCA were not possible due to limited number of studies.

Nonetheless, our analysis provides a number of important outcomes. It is one of the largest meta-analysis addressing an important clinical question. It provides up-to-date data which include both full text papers and abstracts. To date, we lack specific and precise criteria to differentiate between ITB and CD. This is an increasing problem especially with the changing epidemiology of these two conditions worldwide.26, 27 CD is increasing in developing countries, whereby the background prevalence of TB is high; whereas the incidence of TB increasing in the West including in North America is mainly attributable to immigration, human immunodeficiency virus, and multi-resistant TB strains.1 Since IGRA showed good specificity for the diagnosis of ITB, we feel that it has a role in suspected cases before anti-TB therapy is commenced. Furthermore, our data are consistent with a meta-analysis which included five studies.28 Further prospective and large-scale studies should focus on the clinical usefulness when both IGRA and ASCA are used in combination in the diagnosis of ITB or CD.14

In conclusion, our study showed that IGRA and ASCA have high specificity for the diagnosis of ITB. T-SPOT.TB has better accuracy than QuantiFERON-TB Gold, but the result is not based on a head-to-head comparison. IGRA and ASCA may be useful as a supplementary diagnostic tool, especially for patients with unclear initial workup before starting empirical anti-TB therapy.