C-Tb skin test to diagnose Mycobacterium tuberculosis infection in children and HIV-infected adults: A phase 3 trial

Background C-Tb, an ESAT-6/CFP-10-based skin test, has similar sensitivity for active TB compared to tuberculin skin test (TST) and QuantiFERON-TB-Gold-In-Tube (QFT). However, data are limited in children and HIV-infected persons. Methods Asymptomatic South African contacts <5 years (n = 87; HIV-uninfected), or symptomatic individuals of all ages presenting to clinics with suspected TB (n = 1003; 30% HIV-infected) were recruited from eight South African centres. C-Tb and TST were allocated to either forearm double blinded. Samples for QFT were collected in parallel, and test-positivity rates were compared. Results In participants with microbiologically confirmed TB (n = 75; 45% HIV-infected) sensitivity of C-Tb, TST and QFT were similar (72% versus 75% versus 73%; p>0.5). All 3 tests had similar positivity rates in HIV-infected participants with active TB, however, positivity rates were reduced when CD4 counts were <100 cells/μL. In participants where active TB was excluded (n = 920), C-Tb (41%), TST (43%), and QFT (44%) also had similar test-positivity rates. Among asymptomatic contacts aged below five, 32% (28/87) tested positive with C-Tb and 32% (28/87) with TST (concordance 89%). Overall, C-Tb and TST showed a similar safety profile. Conclusion C-Tb was safe and showed similar test-positivity rates, compared to TST and QFT, in children and HIV-infected persons with active or latent M. tuberculosis infection. These data inform the utility of C-Tb in clinical practice. Trial registration ClinicalTrials.gov NCT01642888. EudraCT 2011-005078-40.

group. As there is no microbiological gold standard for LTBI, as in previously published evaluation studies [11,17], sensitivity was determined in active TB [13], with the presumption that sensitivity would be similar or possibly better in LTBI (given that active TB is an immunosuppressive state).

Trial design and participants
The trial was a double-blind, split-body design with randomized injections of C-Tb versus TST in either forearm. The trial was conducted in eight South African sites from September 05, 2012 to September 30, 2014. 1003/1190 of those enrolled were cases with suspected TB (0-65 years; 30% HIV-infected), while the remainder comprised 87 asymptomatic PTB contacts aged below five, and 100 healthy controls (5-11 years old; Fig 1). The latter was to infer the impact of recent exposure on skin test responses using 'supposedly' non-exposed children (no known contact but within a TB endemic environment where the force of infection is high) as a reference, and to complete safety evaluations over a range of childhood subgroups (children with active TB, asymptomatic contacts of active cases, and healthy controls that did not have a definitive history of exposure).
Those with suspected TB between 0 and 65 years of age presented to primary care clinics with TB symptoms (online data supplement S1 Table). These individuals, after investigation and follow-up for 28 days, were classified as either active TB cases (culture or Xpert MTB/RIF positivity served as the reference standard) or non-TB. Participants with clinical symptoms of TB but with a negative microbiology were included in the non-TB group, because only microbiologically confirmed TB was regarded gold standard to evaluate the sensitivity of the tests. The asymptomatic children aged <5 years were close contacts of a smear positive PTB case defined as more than 6 h contact per day for at least five days. The trial staff also recruited a healthy control group of children with no known contact with people infected with M.tb and no symptoms or signs of active TB (as outlined above). Subjects with confirmed active TB or AIDS defining diagnosis at screening were excluded.

Trial procedures
Identical appearing vials of C-Tb and PPD RT 23 (both Statens Serum Institut, Copenhagen, Denmark, (since 2017 PPD RT 23 is a product held by AJ Vaccines) labelled left or right were randomized in blocks of 10, and 0.1 mL of each agent was randomly injected in separate arms immediately one after the other according to the Mantoux technique as previously described [14]. Induration responses for C-Tb and TST were read 48-72 h after intradermal injection transversely to the long axis of the forearm with the ballpoint method and a ruler, and documented with digital images. Indurations !1 mm defined a 'responder'. Samples for Quanti-FERON-TB-Gold-In-Tube (Qiagen, Hilden, Germany) were collected prior to the skin tests. The QFT was performed by PathCare (Goodwood, South Africa) according to the manufacturer's instructions. BCG status was determined according to vaccination cards and the presence of BCG scars. HIV-infection was diagnosed by two alternative positive rapid tests, or one positive rapid test and a confirmatory ELISA. To limit the amount of blood taken from those aged <5 years old (as per ethical approval and guidelines), no blood samples for haematology, biochemistry, QFT or HIV status were collected. The HIV-test result obtained immediately after birth was used.

Safety assessment
Local and systemic adverse events (AE) were assessed by the trial staff through medical assessments 30 minutes after intradermal injection and at the follow-up visits. Diaries were provided for recording of all AEs experienced until Day 28. Injection site reactions (ISRs) were assigned to the respective agent, whereas systemic AEs were assigned to both agents.

Objectives and outcomes
The primary trial objective was to assess the performance of C-Tb in children and HIVinfected persons and to confirm the safety profile. The primary outcome was the test-positivity rate defined by indurations !5 mm for C-Tb and 5 or 15 mm for TST (5 mm if HIV-infected) [10,12,13]. Secondary outcomes were comparisons of C-Tb, TST, and QFT test positivity. As there is no gold standard for diagnosing LTBI, we provided comparative test positivity data for C-Tb, QFT, and TST.

Statistical analysis
Given the lack of a microbiological definition of LTBI, sensitivity was determined in participants with active TB. In those without active TB (non-TB), C-Tb test-positivity rates were compared to QFT and TST. Differences in number of test positives were assessed by McNemar's test for marginal heterogeneity, and in unpaired analyses with Fisher's exact test or χ 2 test. Agreement between tests was assessed with Cohen's κ coefficient. Given that this trial addressed patient safety and used within-person randomisation the results have been presented within the context of the Consort Guidelines. Further details are provided in the online supplement.

Ethics
The trial was conducted according to the principles expressed in the Declaration of Helsinki. Written informed consent form were obtained from adult participants and from parents or legal guardians of children, with additional assent form provided in older children.

Participants and final diagnosis
During the observation period of 28 days following skin testing, active TB was confirmed among 75/1003 (7%) with TB symptoms, while 928/1003 (93%) had no microbiological evidence of active TB (non-TB but this included 44 participants with symptoms who were test negative but empirically treated for TB).
Out of the 928 non-TB cases, two did not receive C-Tb, and six participants were either lost to follow-up, withdrew consent, or died; thus no induration measurements were recorded in these participants. None of the three reported deaths were related to the skin tests. Among the enrolled participants !5 years old, 38 (4%) had missing QFT result, and among the remaining 916, 131 (14%) were indeterminate (all due to low IFN-γ levels in the positive control).
Both genders were equally distributed with the median age of 17 years ( Table 1). The majority of the trial population was of Black African descent (55-87%). The healthy control group inadvertently included participants only of mixed race (91%) and of European descent (9%). There was almost full BCG vaccination coverage among young children. HIV-infection was mainly observed in adults. Participants with unknown HIV status were all 5 years of age. Median CD4 count in HIV-infected adults was 314 cells/μL (IQR 164-502 cells/μL).

Test performance
Overall. C-Tb showed a positivity rate of 40% (95% CI 38-43%, n = 1182), similar to the positivity rate of TST (41% [39-44%], n = 1184, p = 0.3456) and the results were concordant in 84% of the participants. QFT was not done in children <5 years old; for those !5, QFT showed a positivity rate of 43% ([40-46%], n = 916), including 131 (14%) indeterminate results. C-Tb and QFT showed no significant difference among those subjected to dual testing and the results were concordant in 624 (83%) of the 750 participants with a determinate test result (p = 0.1987). The impact of age on test readouts is outlined in the online supplement.
Excluding the 44 participants who were empirically treated for TB from the 920 included in the non-TB group reduced the C-Tb test positive rate by only 2% points (data not shown).

CD4
All three tests showed reduced test-positivity rates in HIV infected participants with CD4 counts <100 cells/μL (Fig 2). This was observed in both those diagnosed with active TB and those without active TB (S5 and S6 Tables). For QFT, the rate of indeterminate results reached a maximum of 65% (17/26) at CD4 counts <100 cells/μL.

Children
In children <5 years old TST results are used to guide treatment for active TB. In the present trial tuberculosis was diagnosed in 12/148 (8%) symptomatic children <5 years old, two of them were confirmed microbiologically and ten diagnosed according to symptoms or signs of TB (S7 Table). The two participants diagnosed with active TB both showed indurations !15 mm with both skin agents, the youngest at four months of age. Both were close contacts with a PTB case and presented with cough >2 weeks and drenching night sweats. Those diagnosed according to clinical symptoms all had TST indurations !14 mm, and all but three tested positive with C-Tb. Of note, 12 children presenting with similar symptoms and positive C-Tb results were not diagnosed with TB (S6 Table). Four of those had no TST induration.
In the asymptomatic group of children with PTB contacts, 38% (33/87) tested positive with C-Tb or TST, 23 being positive with the two tests and five positive with either (p = 0.7518, concordance 89%, Table 4).
The youngest participant testing positive with C-Tb had an induration of 11 mm at 71 days of age, and the youngest responder to TST had an induration of 9 mm at 53 days of age.

Induration size
To explore the impact of age and HIV status on skin test performance we compared the induration sizes of C-Tb and TST responders (defined as skin test induration !1 mm) and the number of non-responders (bimodal distribution). C-Tb showed a minor but significantly decreased induration from median 21 to 19 mm in HIV-infected participants aged 5-65 years (p = 0.0148), whereas TST showed a median of 20 mm irrespective of HIV status (S1 Fig). The number of non-responding active and non-TB participants (5-65 years) increased significantly from 43% (238/550) in HIV-uninfected to 57% (169/296) in HIV-infected for C-Tb, and from 32% (178/551) to 51% (152/296) for TST (p<0.005). Also age had little influence on the  Fig). However, symptomatic children aged below two showed a significantly reduced median of 4 mm (C-Tb, p<0.0001, One Way ANOVA) and 10 mm (TST, p<0.0001) respectively, indicating that indurations were influenced by symptoms in the youngest children ( S2 Fig and S10 Table). Further, the positivity rates for both agents were significantly lower in symptomatic children <5 years old compared to asymptomatic children (Table 4, χ 2 , p<0.05). Of note, the median C-Tb induration among TB symptomatic responders was 20 TST showed a median of 20 mm in participants of Black African ancestry (n = 483), and was significantly reduced to 17 mm in participants of mixed ancestry (n = 128, p = 0.0335), which was associated with the higher BCG vaccination coverage of 85% in the mixed race group compared to 70% in the group with Black African ancestry.

Safety
No serious AEs related to either of the skin tests were observed by the investigators. More than 95% were of mild-moderate intensity. A quarter of the participants experienced at least one ISR from both agents ( Table 5). C-Tb showed similar types and frequencies of ISRs as TST, of which pruritus, pain, and rash were the most common. Injection site reactions appeared most frequently among responders, indicating that the reactions were inevitable parts of the delayed-type hypersensitivity reaction. Hence, adults showed more ISRs than children, and HIV-uninfected more than HIV-infected (not shown). No haematoma were reported with C-Tb.

Discussion
In this phase III trial we found that C-Tb had similar sensitivity compared to the TST and QFT, both in children and HIV co-infected persons with active TB. C-Tb also had similar testpositivity compared to QFT and TST in those with presumed LTBI (i.e. non-TB). These results support the cut-point of C-Tb determined in adults [12,13] may also be applied in children and in HIV-infected individuals. Separate dose-finding and sensitivity trials from the outset (initial phase 1, 2 and 3 trials) would not have been feasible in infants and children due to the paucibacillary nature of TB, their inability to expectorate, and frequent occurrence of extrapulmonary TB [18]. Hence, the need for the current trial in these vulnerable, and often neglected subgroups of patients. Indeed, one of the strategies strongly recommended by WHO to eliminate TB worldwide is systematic testing and treatment of LTBI in people living with HIV, in children, and now also in adolescents and adults, who are in close contact with PTB cases [3,19]. The major potential advantages of C-Tb in endemic countries would be: (i) lack of the need for laboratory facilities equipped with plate readers; these are in short supply and where they exist there is no additional capacity; (ii) lack of the need for additional technicians which are already in short supply; (iii) lack of the need for phlebotomy which is a major hurdle in children; (iv) lack of the need for a transport infrastructure which is already overburdened; (iv) easier test operation by existing health care workers; (v) substantially lower cost (although not yet finalised the tiered pricing for C-Tb will be competitive with TST; current cost of an IGRA to the privately paying end user in South Africa is~US$ 77 and~US$ 10 for the TST [20]).
In keeping with earlier preliminary findings [13], all three tests showed reduced sensitivities in those with CD4 T cell counts below 100 cells/μL, reflecting that the tests relies on competent immune function. The cut-points of TST (5 mm in HIV-infected, 15 mm in others) and C-Tb (5 mm applied in the present trial) were validated in earlier trials [12,13] and in a separate trial that will report a sensitivity of C-Tb similar to TST and QFT (Aggerbeck, in preparation). In the present trial, the positivity rate of C-Tb was similar to TST and QFT except in HIVinfected non-TB participants where C-Tb appeared more sensitive than QFT (but less sensitive than TST) due to a high number of indeterminate results. The reasons for this observation remain unclear but may be related to incubation time with antigen (higher with skin tests) and antigen diversity and load (higher with TST). TST showed the same median induration among HIV-infected and HIV-uninfected responders (induration !1 mm), indicating that a reduced positivity rate among HIV-infected persons is driven by the increased number of nonresponders. These non-responders would have remained test-negative even if a 5 mm cutpoint was applied, however the specificity (63%) would be reduced, as reported previously [8,12,21].
One third of the asymptomatic childhood contacts of PTB aged <5 years old tested positive with C-Tb and TST; it is likely that these children may benefit from LTBI treatment [3]. QFT was not done in this age group and in accordance with WHO recommendations and the reluctance by this group to accept phlebotomy [22][23][24]. Among symptomatic children <5 years old the test-positivity rate of C-Tb and TST was significantly reduced compared to asymptomatic children. Furthermore, the median induration size was reduced, especially in children aged below two. This may be due to the immune-suppressive effect of TB itself, highlighting the need to carefully interpret a negative test result. Consequently, although not investigated in the present trial, in some cases, a repeat test may be considered after 2-3 months on treatment for active TB when the severe illness has resolved [15]. We have previously shown that a repeat test with C-Tb may be feasible [25].
There are several limitations to our trial, including the inherent difficulties in diagnosing active TB in children resulting in misclassification bias (a limitation in all diagnostic trials in children). Indeed, among all paediatric cases reported from 2000 to 2009 in the EU region, only 17% were culture-confirmed [26]. In this trial, QFT had a high percentage (14%) of indeterminate results, resulting in a low test positivity rate in HIV-infected persons. However, a pooled rate of 15% indeterminate results have been reported from high burden countries [27], and high indeterminate rates have been associated with low CD4 counts [28]. Moreover, the test was done by an ISO 15189 certified laboratory, and with a stringent quality assurance system. For example, date and time when samples were collected and received at the laboratory were recorded in addition to incubation conditions as part of the laboratory standard operating procedures. This revealed five cases incubated at an elevated temperature (these results were recorded as missing), and five samples were delayed during transport. It is well known that delays in incubation or a problem in one of the other steps may contribute to the number of indeterminate results [11]. Of note, both skin tests classified one third of the 125 QFT indeterminate results as M.tb infected cases, indicating that simple skin tests may be more suitable than laboratory-dependent tests in resource constrained settings. High QFT sensitivity has been reported in low TB endemic countries [29,30].
We measured sensitivity in active TB. However, as there is no reference standard to diagnose LTBI, sensitivity can only be measured in active TB, and only comparative descriptive statistics (comparing positivity of C-Tb, QFT, and TST) provided in those with non-TB. Alternatively, in the absence of a gold standard, the sensitivity and the specificity of any two tests may be estimated from groups with different prevalence using a latent class model [31,32], although such estimates may be biased if the tests have different sensitivities in the various subgroups as reported here. Moreover, such an approach is still an approximation and can never compensate fully for lack of a gold standard. Although we could not detect any difference in test positivity rates between C-Tb and QFT (excluding indeterminate results), it is possible that small but significant differences could have been missed given the limited sample size. However, the study was primarily powered to detect rare adverse events and not small but potentially significant inter-group differences. Finally, if it is hypothesized that C-Tb (containing RD-1 antigens only) is a more specific test than TST, then positivity rates of C-Tb in the non-TB group should be lower than TST (which was not the case). We believe this is due to the lack of a BCG confounding effect when given at birth. Indeed, a systematic review involving over 240 000 participants enrolled in 24 trials found that the effect of BCG vaccination on TST, when given at birth (the South African practice), was minimal at the 15 mm cutpoint [8,33,34].
There was significant racial disparity in the healthy control group (none were Black African); this inadvertently arose due to the population distribution of the sites where the healthy controls were recruited. However, in all the children (excluding the healthy controls) the C-Tb induration size was similar in those of Black African and mixed ancestry. A higher number of responders (but lower median TST induration) was observed in participants of mixed race, which could reflect the higher BCG vaccination coverage in participants of mixed race (85%) compared to Black Africans (70%) [6].
In conclusion, C-Tb has a similar sensitivity compared to TST and QFT in adults and children with active TB or presumed LTBI using a universal 5 mm cut-point. In HIV-infected persons C-Tb appeared more sensitive than QFT. These informative data could be valuable for clinical practice in the future, suggesting that C-Tb may be a useful tool to guide LTBI-specific treatment as outlined in the WHO End TB Strategy [3,16].   Table. Results of C-Tb, TST, and QFT in 704 active and non-TB participants according to age. Data are presented as n (%). Cut-points defining positive results were 5 mm (C-Tb) and 15 mm (TST). Ã Excluding 299 HIV-infected participants due to an uneven distribution in the various age groups. † Median (IQR) among responders with induration !1 mm. (DOCX) S9 Table. Results of C-Tb versus TST and QFT in healthy control group aged 5-11 years. Ã McNemar's test. Cut-point for TST was 15 mm. In an intention to diagnose principle, QFT indeterminate results were regarded as negative (arrows). † 2 missing QFT. (DOCX) S10