INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: BCG; revaccination; treatment of latent TB infection; tuberculin skin test; tuberculosis

The bacillus Calmette-Guérin (BCG) vaccine is a controversial, yet widely used, vaccine, with an estimated 100 million children receiving it annually throughout the world (1). The majority of these are from countries with intermediate and high incidences of tuberculosis (TB). Singapore is such a country, which has had a national policy of mass BCG vaccination at birth since 1957, and a BCG revaccination program for schoolchildren since 1958. BCG coverage for newborns in Singapore is more than 98%. School entrants at 6 yr of age without a BCG scar or history of BCG vaccination are given direct BCG vaccination by the School Health Service (SHS). BCG revaccination is subsequently administered at 12 or 16 yr of age to tuberculin nonreactors, defined as those with tuberculin skin test (TST) reactions < 10 mm, using 1 TU PPD RT 23 (i.e., 1 tuberculin unit of purified protein derivative of the RT 23 strains). The SHS has utilized 1 TU PPD RT 23 for TST since the 1960s in accordance with the original World Health Organization recommendation (2). No child is given more than two BCG vaccinations, irrespective of the TST reaction. At the age of 12 yr, the majority (about 80%) of Singapore schoolchildren are tuberculin nonreactors, and hence would receive BCG revaccination.

The annual TB incidence in Singapore was about 300 per 100,000 population in 1960, after which it declined steadily to 56 per 100,000 in 1987, remaining at 48 to 56 per 100,000 since (3). This nondecline in TB rates prompted the Singapore government in 1997 to renew its political will and commitment toward TB control in the form of the Singapore TB Elimination Programme (STEP) (4). Key elements of STEP which exist in addition to TB control program efforts are the surveillance of clinical progress and treatment completion rates of all patients receiving TB treatment in the country, and the implementation of directly observed treatment (DOT) at government family health clinics island-wide. Another major intervention under STEP is the active tracing and screening of close contacts of all infectious TB cases to identify candidates for treatment of latent TB infection. Before 1998, treatment of latent TB infection had been offered only to children who were close (household) contacts of TB cases, and to schoolchildren discovered on routine TST screening for BCG revaccination to have TST  18 mm.

Since its introduction more than a century ago by Robert Koch, the TST has remained the only practical means by which to identify persons infected with Mycobacterium tuberculosis. However, infection with mycobacteria other than TB (MOTT) and previous BCG vaccination also cause reactions to tuberculin (5, 6), thus confounding the interpretation of the TST. The TST cutoff readings recommended by the American Thoracic Society (ATS)/Centers for Disease Control and Prevention (CDC) to identify candidates for preventive treatment of latent TB infection apply to non-BCG-vaccinated populations (7, 8). A cutoff reading of  5 mm is recommended by the ATS/CDC to be considered as positive in household contacts of a tuberculosis patient (8). The United States Advisory Council for the Elimination of Tuberculosis (ACET) has recommended that the treatment of latent TB infection be considered for any BCG-vaccinated person with a TST reaction of  10 mm, especially if the person is a contact of an infectious TB case, or if the person was born or has resided in a country of high TB prevalence (9). This recommendation as it pertains to those who received BCG vaccination in infancy is reasonable; however, in persons vaccinated after infancy, it has been shown that higher TST readings are common (6). It should also be noted that the cutoff TST readings recommended by the ATS/CDC and ACET are based on 5 TU of PPD-S, which is the bioequivalent of 2 TU of PPD RT 23 (10, 11).

To our knowledge, there are few data to support the choice of TST cutoff reading on which to base the recommendation for treatment of latent TB infection for the various risk groups in a BCG-vaccinated or revaccinated population. As the majority of the Singaporean population born after the mid-1950s would have received two BCG vaccinations, we had to empirically choose higher TST cutoff points than that recommended by the ATS/CDC and ACET to guide our decision for treatment of latent TB infection in each of the different risk groups.

As there are School Health Service screening as well as TB notification data available for several birth cohorts of Singapore schoolchildren who were tuberculin tested at the ages of 12 and 16 yr, we were able to analyze the risk of developing TB disease according to TST reading in these birth cohorts. We sought to determine the TST cutoff readings that would be the most valid in predicting latent TB infection in these low-risk groups of schoolchildren with prior BCG vaccination/ revaccination. In the absence of similar data on higher risk groups (e.g., close contacts) with prior BCG vaccination/revaccination, we reasoned that it would be acceptable (or at worst, erring on the conservative side) to extrapolate our findings to guide the decision on whether to initiate treatment of latent TB infection in higher risk categories.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This nonconcurrent cohort study merged data from two separate computerized databases: Data on exposure variables including TST reading were extracted from the School Health Service database. The outcome of TB disease was established from notifications to the National TB Registry. Correct cross-matching of individual records was assured by the use of names and, for long-term Singapore residents, National Registration Identity Card numbers.

The Singapore School Health Service has kept computerized records since 1990. A database of TST readings at 12 yr of age for school cohorts born from 1978, and at 16 yr of age for cohorts born from 1974, therefore exists. These records are archived once each cohort reaches 21 yr of age. Hence, at the time of this study in 1998, records of cohorts born before 1977 were unavailable for analysis. We therefore extracted for analysis individual records of TST readings at 12 and 16 yr for birth cohorts from 1978 to 1984, corresponding to the primary 6 (sixth-grade) cohorts of 1990 to 1996, and the secondary 4 (tenth-grade) cohorts of 1994 to 2000. Birth cohorts from 1985 on were omitted because at the close of the study period they would have experienced less than 1 yr at risk for development of TB disease after tuberculin skin testing at 12 yr of age in 1996.

Occurrences of TB in the study cohort were established through the Singapore National TB Registry, which was set up in 1957, with notification of TB cases mandated by law. All episodes of TB occurring up to 1997 within the age range 13 to 16 yr among the birth cohorts from 1978 to 1984representing the 4 yr after the first TST reading in primary 6were identified from the National TB Registry and related to the pertinent records in the database of TST readings. Using person-time analysis, the 4-yr cumulative incidence of TB disease was first analyzed by TST reading in categories of 0-4, 5-9, 10-14, 15-17, and  18 mm. These demarcations are meaningful because they represent frequently proposed cutoff decision points used in the management of latent TB infection. The rate ratio (RR), using 0-4 mm as the base category, was determined to elucidate the magnitude of increase in risk of TB disease with increasing TST size category. A receiver-operating characteristic (ROC) curve was constructed to ascertain the ideal TST reading cutoff for prediction of TB disease in the population aged 13-16 yr.

Similarly, notifications of TB up to 1997 within the subsequent age range of 17 to 20 yr among the birth cohorts of 1978 to 1980 were identified from the National TB Registry and related to the TST reading database. The 4-yr cumulative incidence and RRs of TB disease were analyzed by TST reading in categories of 0-4 mm (base) and 5-9, 10-14, 15-17, and  18 mm. An ROC curve was constructed to ascertain a suitable TST reading cutoff for prediction of TB disease in the population aged 17-20 yr.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

There were 266,005 children in the birth cohorts of 1978 to 1984 with TST readings available at 12 yr of age. Of the total, 82.8% (n = 220,278) were tuberculin nonreactors (TST reading < 10 mm). These nonreactors received BCG revaccination, whereas the remaining 17.2% (n = 45,727) who had TST  10 mm did not. Of the latter, 41,028 children had TST readings between 10 and 14 mm, and 4,699 readings  15 mm.

Fifty-nine episodes of TB occurred between the ages of 13 and 16 yr in the study cohort. The cumulative incidence of TB disease during this 4-yr follow-up period, stratified by TST reading at age 12 yr, demonstrated a clear and statistically significant trend of increase with TST size (p < 0.001, trend test) (Figure 1).

View larger version (35K): [in this window] [in a new window]   Figure 1.   The cumulative incidence of TB disease during the 4-yr follow-up period after TST at age 12 yr, stratified by TST reading at age 12 yr. This demonstrates a clear and statistically significant trend of increase with TST size (p < 0.001, trend test).

The RR of TB disease was greatest in the category  18 mm (RR, 47.9, 95% confidence interval [CI], 19.1-120.1; Table 1). The ROC curve indicated a TST reading of 10 mm to be the most sensitive and specific cutoff for predicting the development of TB disease in this age group (Figure 2).

View larger version (17K): [in this window] [in a new window]   Figure 2.   The ROC curve indicates a TST reading of 10 mm to be the most sensitive and specific cutoff for predicting the likelihood of developing TB disease in the 4 yr following TST at age 12 yr in schoolchildren who had received only one BCG vaccination (at birth). The diamonds indicate the various TST cutoff values (greater than and equal to).

In the cohort analysis of the next period, there were 122,485 children from the birth cohorts of 1978 to 1981 with TST readings available at 16 yr of age. Of these, 36% (n = 43,906) had TST < 10 mm; 49% (n = 60,104) had TST between 10 and 14 mm, and 15% (n = 18,475) had TST  15 mm.

There were 76 TB cases in the 4-yr follow-up period between the ages of 17 and 20 yr for this cohort. In person-time analysis, the cumulative incidence of TB disease, stratified by TST reading at age 16 yr, again demonstrated a clear and statistically significant trend of increase with TST size (p < 0.001, trend test) (Figure 3).

View larger version (33K): [in this window] [in a new window]   Figure 3.   The cumulative incidence of TB disease during the 4-yr follow-up period after TST at 16 yr, stratified by TST reading at age 16 yr. This demonstrates a clear and statistically significant trend of increase with TST size (p < 0.001, trend test).

The RR of TB disease was greatest in the category  18 mm (RR 18.3, 95% CI, 7.2-46.9; Table 2). The ROC curve supported a TST reading of 16 mm as the most sensitive and specific cutoff for predicting the development of TB disease in this age group (Figure 4).

View larger version (17K): [in this window] [in a new window]   Figure 4.   The ROC curve supports a TST reading of 16 mm as the most sensitive and specific cutoff for predicting the likelihood of developing TB disease in the 4 yr following TST at age 16 yr, the majority (about 80%) of whom had received BCG revaccination at age 12 yr. The diamonds indicate the various TST cutoff values (greater than and equal to).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The interpretation of the TST in a BCG-vaccinated individual often poses a dilemma for the TB control health worker. Studies have shown postvaccination BCG-induced tuberculin reactivity to range from no induration to reactions of 19 mm (12). Factors affecting the size of the tuberculin reaction after BCG vaccination include the vaccine manufacturer (14), dose (15), and method of administration (16). Tuberculin reactivity to BCG given in infancy wanes rapidly within 5 yr (12, 16). However, those vaccinated at a later age have been shown to demonstrate larger TST reactions that wane more slowly, with 15 to 25% persisting beyond 10 yr (6, 14). It has been shown in this subgroup that the size of the reaction may be similar to that in TB-infected persons (6). BCG-induced reactivity that has waned may be boosted.

Menzies and coworkers have shown that, among non-Canadian/U.S.-born schoolchildren and young adults in Montreal, history of BCG vaccination appeared to be an important cause of reaction in subjects from low-incidence countries but diminished in importance with increasing incidence of TB in their country of origin (20).

Our study examined the performance of the TST, using 1 TU of PPD RT 23, as a screening test for latent TB infection in a teenage population in Singapore. By measuring the subsequent incidence of TB disease in a cohort of Singaporean schoolchildren born between 1978 and 1984 over two consecutive age intervals, from 13-16 and 17-20 yr, we established the predictive value of TST readings taken immediately before each interval of follow-up, at ages 12 and 16 yr, respectively.

Given only a single TST reading at the beginning of each 4-yr period, some misclassification of exposure is inevitable in our study, becoming more common with increasing remoteness from the time of skin testing. Accordingly, it is possible that some cases of TB disease progressed from infection acquired after the time of tuberculin skin testing in each period: these would have unexpectedly low TST readings. The upward trends in incidence with increasing TST size for both age groups in our study emerge clearly despite this bias towards the null (Figures 1 and 3).

An alternative explanation for this finding would be the protective effect of BCG revaccination. We have assumed that this was not the case, as there is no available evidence thus far to support the efficacy of BCG revaccination in protecting against TB disease (21).

Despite an existing policy of offering treatment of latent TB infection to schoolchildren with TST  18 mm during the study period, only 119 schoolchildren (representing 3% of those with TST  18 mm from 1990 to 1997) received treatment for latent TB infection. Their compliance and treatment completion rates are not known. The possibility that this intervention may have confounded our analysis is hence almost negligible.

The children who were tuberculin tested at age 12 yr were all assumed to have received one BCG vaccination, with virtually all having received it at birth. In the population that was tuberculin skin tested at age 16 yr old, we were unable from available data to distinguish those who did and did not receive BCG revaccination at age 12 yr. However, according to School Health Service records, more than 80% of these students were tuberculin negative at age 12 yr, and would therefore have been BCG revaccinated. Of these students, most of whom would have received a second BCG vaccination within 4 yr of TST, larger TST readings at age 16 yr would not be unexpected, due possibly to the effect of the second BCG vaccination. This was indeed the case, with 64% of our schoolchildren having TST  10 mm at age 16 yr old versus 17.2% at age 12 yr old.

This study supports a recommendation of a TST reading of  10 mm (using 1 TU of PPD RT 23) as a cutoff to indicate an increased risk of development of TB disease, in the screening of Singaporean schoolchildren who had received BCG vaccination at birth. However, in older children with two previous BCG vaccinations, our findings support using a TST reading cutoff of  16 mm.

Our findings raise the question as to the feasibility of recommending treatment of latent TB infection in this group of low-risk persons without any recent tuberculosis exposure at the cutoff points mentioned above, according to the number of previous BCG vaccinations received. Whether it would be cost-beneficial (in terms of per case of TB prevented, and resource prioritization and utilization) to recommend treatment of latent TB infection for Singaporean schoolchildren found to have TST  10 mm at age 12 yr, and TST  16 mm at age 16 yr, is an issue that deserves further study.

These findings in a low-risk group have also provided guidance in the formulation of criteria for identifying those among higher risk groups (e.g., close contacts) with prior BCG vaccination/revaccination to whom it would be appropriate to offer treatment for latent TB infection when they present at the tuberculosis control unit for screening. In extrapolating this result, we noted that the expected predictive value of a positive TST in such a higher risk population would be even higher than in our study population.

The appropriateness of any such strategy based in part on TST readings must be reviewed with reference to existing TB prevalence rates in the community. With future decreases below the current intermediate level of TB prevalence in the Singaporean population, the rising proportion of false positives who would receive unnecessary preventive treatment will alter the relative balance of benefit against risk and cost. Other factors in this decision would be the effectiveness of treatment of latent TB infection, and the likelihood of adherence to and completion of treatment.