1.1. Study objectives

Study objectives were firstly to estimate rates of QuantiFERON-TB Gold In-Tube (QFT) conversions and reversions in Indian health care trainees, during six-month intervals, over four test points. Secondly, rates of TB exposure were estimated by type of health care trainee, and evaluated whether QFT conversions were associated with TB exposures, and QFT reversions with treatment. HCWs were categorized by patterns of QFT responses, to assess whether different patterns were associated with exposure. Secondary objectives were to assess what occupational TB exposure factors (if any) were associated with change in IFN-gamma response over the four time points.

2.1. Study setting and population

This study was conducted at the Mahatma Gandhi Institute of Medical Sciences (MGIMS), Sevagram, a teaching hospital located in central India. The hospital sees an average of 300 smear positive TB cases per year, two thirds of which are treated as in-patients. As MGIMS is a referral center, most smear positive patients are seriously ill when they reach the hospital and require hospitalization. Approximately three quarters of all smear positive in-patients are diagnosed only during in-patient evaluation.

TB infection control measures are limited, screening for LTBI is not routinely performed and personal respirators are not routinely provided (and when available not routinely employed by HCWs when interacting with patients with active TB). Previous research at this hospital has shown a high prevalence of LTBI, high annual risk of TB infection in young health care trainees, and a high incidence of active TB among HCWs [7,15,18].

2.2. Study design and ethics approval

This longitudinal study was conducted from August 2007 to July 2009. Upon providing informed consent, students completed a brief questionnaire on occupational and non-occupational exposure to TB prior to enrollment, and blood was drawn for the QFT. Participants were followed up for 18 months and underwent blood draws and exposure assessment every 6 months. TST was not done as TST could have altered QFT responses [19]. Ethical approval for the study was obtained from both the McGill University Health Centre (MUHC) in Montreal, Canada, and from the institutional review board (IRB) of the MGIMS.

2.3. IGRA testing (QFT)

Participants were tested with the QuantiFERON-TB Gold In-Tube (QFT) test (Cellestis, Inc./Qiagen, Carnegie, Australia), performed as per the manufacturer’s instructions on site at the MGIMS laboratory. The QFT was considered positive if the interferon-gamma (IFN-g) produced in the TB antigen tube minus the IFN-g produced in the nil tube was ⩾0.35 IU/ml, as per the manufacturer’s recommendations. Quantitative IFN-g values were also recorded; results greater than 10 IU/ml or less than 0 were truncated at 10 and 0 respectively; as values outside this range are not precisely measurable.

2.4. Questionnaire and ascertainment of TB exposure

Exposure ascertainment was done by verifying detailed log books kept by all trainees, outlining the students’ daily responsibilities between each study test point. Log books are used to monitor attendance during the students’ hospital rotation duties, and include what ward and patients (using a 12-digit hospital ID number) individual students were responsible for each day. Typically, medical students took part in 60-min rounds, while nursing students had 15-min interactions per patient, but had a higher number of patient interactions per day, in which they help with vital signs, medications, and bed care. Electronic discharge summaries of all patients who interacted with HCW trainees were obtained to identify if they had TB, and it was noted if it was pulmonary (sputum smear positive or smear negative) or extra-pulmonary, and whether the patient was on anti-tuberculosis treatment (ATT) at the time of contact. In summary, the measurement of exposure (i.e., quantity and duration of contact) for each individual student participating in the study was based on actual time spent caring for infectious TB patients. Occupational TB exposures prior to study enrollment and non-occupational TB exposures were collected through self-administered questionnaires completed by the HCW at each test point.

Using these detailed exposure durations collected for each student, three cumulative exposure variables were constructed: (a) total number of days exposed to smear positive or smear negative TB cases; (b) actual minutes exposed to smear positive TB cases only; and (c) number of days exposed to an active TB patient, not on ATT.

2.5. Study definitions

Both conventional and alternative QFT conversion definitions were assessed. Conversion definition (A) uses the manufacturer’s recommended cut-off point of a change from negative to positive (IFN-g ⩾ 0.35 IU/ml). Conversion definition (B) was a change from negative to positive (cut-off point IFN-g ⩾ 0.35 IU/ml) and an absolute increase of 0.7 IU/ml over the first QFT result; this comes from within-subject variability research conducted in South Africa [19,20]. Conversion definition (C) was the most stringent, with a change from negative to positive based on a threshold value of IFN-gamma ⩾ 1.0 IU/ml, reflecting earlier research conducted by this study’s group [21]. QFT reversion was defined as a change from a positive result to a negative result (cut point <0.35 IU/ml). Conversions and reversions were assessed with reference to the QFT result from the previous testing period (i.e., 6 months earlier).

Based on previous work, this research expected to find five patterns of changes in dichotomous QFT results over time [22]. The five patterns were: (a) persistent negatives – all QFT negative results; (b) persistent positives – all positive QFT results; (c) stable conversions – QFT conversion at any time point, and no subsequent reversion; (d) stable reversions – QFT reversion at any time point, with no subsequent conversion; and (e) unstable patterns – at least one conversion and one reversion over the study period [22].

2.6. Statistical analysis

All analyses were performed with Stata version 11 (Stata Corp, Texas, USA). The various definitions listed above were used to assess rates of QFT conversions and reversions. To evaluate the association between the patterns of QFT results and cumulative TB exposure, odds ratios and their accompanying 95% confidence intervals (CI) were estimated from the multivariable polytomous logistic regression model, which allows for the simultaneous assessment of multiple outcome categories [23].

Multiple longitudinal modeling techniques were used to assess an association between recent occupational TB exposure and one of three outcomes: QFT positivity, conversions, or continuous IFN-gamma. General estimating equations (GEE) were used to estimate population-averaged marginal models for QFT positivity [24]. It was investigated whether there was a difference in the effect of the time dependent occupational exposure variables within-subject versus between subjects over time [25]. Generalized linear mixed models (GLMMs), estimated via adaptive quadrature, were used to assess whether time-dependent occupational exposures were associated with QFT conversions, using one of the three proposed conversion definitions. Random intercepts and random slopes were investigated for test point, initial interferon-gamma values and key time dependent occupational exposure variables. The final model comparing QFT conversions (definition A) was a random intercept only model, dummy variables were used to account for high moderate and low initial interferon gamma values, an interaction term was investigated for test point and initial IFN-gamma response at study enrollment. Finally, a linear random effects model was fit to continuous IFN-gamma values. Given that continuous IFN-gamma has both floor and ceiling effects, due to the truncation of continuous IFN-gamma values at 0 and 10 IU/ml, there was concern about the assumption of normality. Therefore, a mixed effects Tobit regression [26] was used to assess the association between exposure and rates of change in continuous IFN-gamma values while accounting for the ceiling and floor effect seen in the distribution of continuous IFN-gamma.

3.1. Participant demographics and exposure to TB

A total of 270 trainees (165 [61.1%] medical and 105 [38.9%] nursing students) participated in the study and completed testing. Table 1 summarizes the characteristics of the study cohort. A total of 61.5% were female, median age was 19.4 years. Valid QFT results were available for all 4 time points in 196 participants, while 28 participants completed 3 tests, and 46 participants completed 2 tests (a total of 960 valid QFT results). Based on the log books maintained by trainees, participants reported a median number of 24, 25.5 and 12 days spent on medical wards between each test point respectively; a visual representation of time spent on the wards for each interval is presented in Figure 1. Over the course of the study, 52 participants (19.3%) were advised to take INH based on positive test results at study enrollment. Among them, 18/52 (34.6%) participants initiated INH therapy, and all 18 completed 6 months of therapy. No participant was diagnosed with active TB during the study period.

Figure 1

Days spent on medical wards for each interval between testing. Histogram of number of days spent on medical wards for each student during each 6 month interval between QFT testing. Frequency is the number of participants with that number of exposed days per testing period.

Upon initial testing, 64/226 (28.3%) HCWs reported positive QFT test using the manufacturer’s definition.

Table 2 describes the mean number of occupational exposures per testing period stratified by type of trainee (i.e., medical or nursing). With the exception of TB exposure in the community prior to study enrollment, nursing students report consistently a higher mean number of days exposed during each testing period. While medical students reported a mean of 15.5 (SD: 15.1) days working on medical wards per 6 month period, nursing students reported a mean of 80.9 (SD: 75.9) days. The only exposure variable to be more common in medical students was non-occupational TB exposure prior to study enrollment, which was reported in 24/420 (5.7%) nursing students and 68/660 (10.3%) medical students. QFT positivity at study enrollment was similar between groups; 27.5% (44/160) QFT positive medical students and 20/66 (30.3%) QFT positive nursing students.

3.2. Conversion and reversion rates

When QFT conversion rates were estimated among medical and nursing students, treating each period as an independent observation, the QFT conversion rate (54/307) among medical students was 17.6%, and the QFT conversion rate (9/134) among nursing students was 6.7%, despite higher levels of reported exposures in nursing students. Employing a more stringent cut-off point consistently resulted in decreased conversion rates across all test points, but little difference was found between the more strict conversion definitions (B and C).

Table 3 presents the total number of conversions, at any test point, stratified by participant characteristics and TB exposure at enrollment. Conversions were less frequent among participants with occupational and non-occupational TB exposures prior to study enrollment. For occupational exposures collected post-study enrollment, the mean number of days or minutes exposed per 6 month period between testing dates was calculated (Table 4). The Mann–Whitney test was used to assess whether the median number of exposures was different between participants with conversions and participants with stable negatives. For mean number of days spent working on medical wards, it was found that converters had a statistically smaller mean number than did stable negatives (17.1 vs. 28, p = 0.01). Similarly, converters reported a smaller number of mean days in contact with TB patients not on anti-TB therapy (0.71 vs. 1.6, p < 0.01).

Conversion and reversion rates for each test period are presented in Table 5. Using the manufacturer’s conversion definition, 18/159 (11.3%) conversions were identified at QFT 2 among participants with negative QFT 1. At QFT 3, 16/143 (11.2%) conversions were identified among QFT 2 negative participants, and 29/139 (20.9%) conversions at QFT 4 were found among participants with negative QFT 3 results. Reversion rates were frequently higher than conversion rates, 16/61 (26.2%) at QFT 2 among those with positive QFT 1 results, 16/63 (25.4%) at QFT 3 among participants with positive QFT 2 results and 12/60 (20%) reversions at QFT 4 among participants with positive QFT 3. Denominators for reversion rates are smaller at each test point compared with conversion rates as fewer participants were QFT positive and thus eligible for a reversion upon subsequent testing.

3.3. Conversions and/or reversions stratified by initial IFN-gamma values

Conversion and reversion rates using the manufacturer’s cut-off point and stratified by initial quantitative QFT results are presented in Table 6. The majority of these trainees were young and had limited exposure to TB prior to enrollment in the program, therefore IFN-gamma response at study enrollment might be a useful predictor of QFT patterns over time. Conversion and reversion rates were calculated at each test point, the denominators for conversion/reversions rates were all HCWs with a negative/positive QFT at the previous test point. Therefore, a HCW with an initial QFT 1 positive could have a reversion at QFT 2 and a subsequent conversion at QFT 3. Higher conversion rates at QFT 3 and QFT 4 can be seen among those participants with higher QFT 1 IFN-gamma results (i.e., 33–80% conversion at QFT 3 among QFT 1 positives vs. 9–19% QFT 3 conversion rates among those with QFT 1 negative results). Similar patterns can be seen among QFT reverters, where among QFT 1 negatives, reversion rates were high and ranged from 67% to 80%, while among the QFT 1 positives, reversion rates were between 0% and 33%. This was investigated further in the longitudinal analysis.

3.4. Patterns of change over 18 months

Each participant was classified into one of five groups based on their cumulative 18 month QFT pattern. Using the manufacturer’s definition of a positive test, 151/270 participants (55.9%) were persistently negative, 45/270 (16.7%) were persistently positive, 29/270 (10.7%) participants had a stable conversion, 11/270 (4.1%) had a stable reversion, and finally 34/270 (12.6%) of participants had unstable patterns. Figure 2A–E are trajectory plots of the quantitative IFN-g values over time within each participant stratified by QFT patterns. These demonstrate the wide variability of responses over time within some participants, and the diversity across subgroups. The variability among persistently positive and unstable patterns appears to be much larger compared with the variability seen among the persistently negatives. These patterns should be interpreted as a slice in time; looking at Figure 2 it can be observed that many of the stable converters converted at QFT 3 or QFT 4.

Figure 2

Trajectory plot of interferon-gamma (TB antigen – nil) over 4 test points, by participant. Trajectory plots of continuous interferon-gamma (TB antigen – nil, IU/ml), within each participant over time. Plots are presented for all students with the same 18 month QFT trajectory together (i.e., A: all students with persistent negative patterns; B: all students with persistent positive patterns, etc.).

Multivariable polytomous logistic regression was performed to assess whether known TB risk factors and cumulative 18 month occupational exposure were associated with any particular QFT pattern. Stable converters were more likely to have had exposure to TB in the hospital prior to baseline testing (OR = 14.3, 95% CI: 2.1–96.8), while unstable patterns were significantly less likely to have had exposure to TB in the hospital prior to study enrollment (OR = 0.2, 95% CI: 0.0–0.8).

3.5. Longitudinal analysis: QFT positivity and conversions

Next, it was assessed whether (a) QFT positivity or (b) QFT conversions (both binary outcomes) were associated with occupational exposures to TB occurring between each QFT test point. Occupational exposure variables were collected in a time-dependent manner (i.e., for each interval between testing) and included: days/time working on medical wards, days/time exposed to smear positive patients, total days/time exposed to smear positive and smear negative and extra-pulmonary patients, and finally total days exposed to active TB patients not on ATT. Non-linear shapes for the exposure outcome association were investigated, but the best fit was a linear characterization of exposure.

3.6. Longitudinal analysis: change in continuous interferon-gamma

To evaluate whether exposure over time was associated with rates of change in continuous IFN-gamma values, a random effects Tobit model was used to account for the ceiling and floor effect seen among continuous IFN-g values, which are truncated at 0 and 10 (Figure 3). In univariable analysis using the Tobit model, it was found that days having worked on medical wards (within-subject effect) were statistically significantly associated with a positive change in quantitative IFN-gamma, but this was no longer statistically significant in the final multivariable Tobit model. High initial IFN-gamma values (>1.0 IU/ml) were associated with a 6.09 unit increase in IFN-g over time in the adjusted model (95% CI: 5.08–7.10).

Figure 3

Histogram of continuous interferon-gamma results from all time points. Continuous interferon-gamma (TB antigen minus the nil) measured using the QFT-GIT test, from all participating students and across all time points, n = 960.

4.1. Strengths and limitations

A major strength of this study was its being conducted in a high TB incidence setting. There are few serial testing studies from high TB incidence settings, and fewer still that have examined any sort of occupational exposure. Another key strength of this work was the direct ascertainment of detailed exposure to infectious TB patients. The study design took advantage of detailed records of occupational and community TB exposure within the medical school training program to directly capture actual time spent caring for active TB patients. In addition, by using discharge summaries, the smear status of active TB cases and whether patients were on anti-tuberculosis treatment at the time of contact were identified, thus establishing whether patients were contagious at the time of contact. Furthermore, all medical and nursing trainees at this school are required to live in on-campus dormitory-style housing, so that community exposure during their training program is relatively low.

However, India is a high TB incidence setting, and the possibility of unknown or un-documented exposure to TB cannot be ruled out. While a major attempt was made to obtain detailed exposure assessment, the possibility that the log books which are used for attendance and monitoring are not entirely accurate cannot be eliminated, any extracurricular activities or additional time trainees spent in the wards would not have been recorded as part of the training logs.

Finally, modeling techniques particularly suited to longitudinal data were used to account for the within-subject serial correlation of QFT results over time. However, these although a large number of test points were available, the small number of outcome events, together with a large number of potential predictors, means that confidence intervals are wide, and inference should be made with caution.