Study population and clinical characteristics
A total of 3729 subjects with suspected diagnosis of pulmonary TB were recruited from three tertiary hospitals. Finally, there were 2390 (64.09%) patients with positive T-SPOT results enrolled in the study. The patients(n=10) refused to receive diagnostic anti-TB treatment, and the patients(n=11) lost follow-up during anti-TB treatment. Among these patients, 1549 (64.8%) patients were finally diagnosed with ATB, including 1091(70.4%) confirmed TB cases, and 458 (29.6%) probable ATB (Fig. 1).There were 724 (30.3%) patients were finally diagnosed with non-TB. Among the 724 non-TB cases, including 39.2% (284/724) bacterial pneumonia,16.3%(118/724)bronchiectasis,12.3%(89/724)lung cancer, 11.8% (86/724) nontuberculous mycobacteria infection,7.3% (53/724) fungal infection, 4.6% (33/724) virus infection and 1.9% (14/724) parasitic infection; and 6.4% (47/724)other diseases. 96 (0.04%) patients without final diagnosis (Figure 1). The demographic and clinical characteristics of the participants are shown in Table 1.
No significant difference was found in sex, human immunodeficiency virus infection, hematologic malignancy, solid tumor, chronic renal failure, autoimmune disease receiving treatment, diabetes mellitus. The rate of history of TB in the ATB group was higher than that in the non-TB group. More liver cirrhosis in ATB group than non-TB group(P=0.016),and transplantation receiving treatment( P=0.036)( Table 1).
Diagnostic performance of the T-SPOT.TB assay and TBAg/PHA ratio for active TB
We observed a significant increase in the results of ESAT-6in ATB group compared with non-TB(P<0.0001) (Fig 2A, Tab 2). The mean value of ESAT-6 were 90±107.2 (mean ± SD) sfc in ATB group, and 16.9±23.91 sfc in non-TB group. The mean value of CFP-10 were 150.1±161.7 sfc in ATB group and 18.48±27.59 sfc in non-TB group. There was also a significant increase in the results of CFP-10 in ATB group compared with non-TB(P<0.0001) (Fig 2A, Tab 2) .
Compared the results of ESAT-6 in Confirmed ATB group(mean ± SD:92.67±109.6 sfc) with non-TB group, and Probable ATB group(83.63±101 sfc) with non-TB group, there were statistically significant differences(both P<0. 001). Compared the results of CFP-10 in Confirmed ATB group(mean ± SD:160±166.8 sfc) with non-TB group, and Probable ATB group(126.4±146.1 sfc) with non-TB group, there were statistically significant differences(both P <0. 001).
There was a significant decrease in the results of PHA in ATB group compared with non-TB(P<0.0001) (Fig 2A, Tab 2). The mean value of PHA was 286.3±165.5 (mean ± SD) SFCs/2.5×105 in ATB group, and 387.1±152.2 SFCs/2.5×105 in non-TB group. Compared the results of PHA in Confirmed ATB group(273.9±163.3 SFCs/2.5×105) with non-TB group, and Probable ATB group(315.9±167 SFCs/2.5×105) with non-TB group, there were statistically significant differences(both P <0. 001).
The AUC of the receiver operating characteristic (ROC) curve for ESAT-6 was 0.832 (95% CI 0.814–0.849) in distinguishing ATB from non-TB, CFP-10 was 0.839 (95% CI 0.823–0.855), and PHA was 0.671(95% CI 0.648–0.694) (Fig 2B, Tab 2). Meanwhile, we compared the confirmed ATB and Probable ATB vs non-TB, the results were similar (Fig 2A, Tab 2).
We calculated the TBAg/PHA ratio of every individual (the larger value of ESAT-6 sfc to PHA sfc or CFP-10 sfc to PHA sfc). We observed a significant increase in the results of TBAg/PHA ratio (P<0.0001) in ATB group compared with non-TB group(Figure 2B). The median values and IQR of TBAg/PHA ratio were 0.5(0.192-1.10) and 0.052(0.029-0.097) in the ATB group and the non-TB group. The AUC of ROC curve for TBAg/PHA ratio was 0.911 (95% CI 0.899–0.922) in distinguishing ATB from non-TB(Figure 2B).
These data suggest that directly using PHA results has no enough value(AUC =0.671) in distinguishing the ATB and the non-TB. The AUC of ESAT-6 and CFP-10 were performed a certain value in distinguishing the two group, but the TBag/PHA ratio showed outstanding performance. Therefore, we focus on a comprehensive analysis of the TBag/PHA ratio.
The performance of different thresholds of Tbag/PHA ratio for ATB diagnosis
According to the statistical analysis of ROC curve, the TBag/PHA ratio in distinguishing the ATB group and the non-TB group showed that the best threshold value was 0.1443. When the threshold value of 0.1443 was used, the sensitivity was 81.73% (95% CI 79.71% -83.63%%)and the specificity was 86.6% (95% CI 83.9%- 89%), giving a PPV of 92.9 (95% CI 91.5-94.0%) and a PLR of 6.1 (95% CI 5.1-7.3) for ATB. This optimal threshold (0.1443) may show the best balance between sensitivity and specificity, but we generally need better specificity and sometimes need better sensitivity. Therefore, we calculated the performance of different thresholds, as shown in Table 3.
According to ROC analysis, if a specificity of more than 90% will be obtained, TBag/PHA ratio threshold value of 0.1817 should be used to discriminate between ATB and non-TB. When TBag/PHA ratio threshold value was 0.2004, the specificity was 92.27% (95% CI 90.07%- 94.1%) and the sensitivity was 74.37% (95% CI 72.12%- 76.53%) (Table 3).If sensitivity of more than 90% will be obtained, a threshold value of 0.0783 should be used, the sensitivity was 90.06% (95% CI 88.46%- 91.5%) and the specificity was 67.82% (95% CI 64.28%- 71.21%). Compared TBag/PHA ratio in the confirmed ATB group or probable ATB group vs non-TB group, the ROC analysis results were similar(Figure 2B, Table 3).
The TBAg/PHA ratio and diseases affecting host immune status
We analyzed the relationship between the results of T-SPOT.TB assay and diseases affecting host immune status. The underlying conditions listed in Table 1 were considered as immunosuppressive status, such as HIV infection, autoimmune disease receiving treatment, etc. The results of ESAT-6 and CFP-10 in immunocompromised patients were lower than others, while ESAT-6 and CFP-10 results in non-TB patients were even lower(Figure 3A). But the results of PHA were different, PHA value increased in non-TB patients (Figure 3A). The result of TBAg/PHA ratio showed that it was the highest among immunocompromised patients. ROC analysis showed that compared with ESAT-6 and CFP-10, the AUC of Tbag/PHA ratio in immunocompromised ATB and non-TB was closer, suggesting that Tbag/PHA ratio is less affected by the immune status of the host (Figure 3B).
The TBAg/PHA ratio and the types of ATB and patient's condition
Some types of tuberculosis or the patient's condition shows that their immune status was different from that of general ATB patients, T-SPOT assay's results can reflect it. We observed that ESAT-6 sfc, CFP-10 sfc, and TBAg/PHA ratio were increased in miliary tuberculosis patients and drug-resistant ATB patients compared with all ATB patients (Figure 4A). But it was not observed in patients with cavitary pulmonary tuberculosis(cavity ATB) (Figure 4A). Furthermore, we observed that the TBAg/PHA ratio was positively correlated with erythrocyte sedimentation rate(ESR), C-reactive protein, and peripheral blood adenosine deaminase in ATB patients, while it was negatively correlated with the percentage of lymphocytes (Figure 4B). However, we did not observe a significant correlation between TBAg results with these indicators (Figure 4).