INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The variable and often long latency period between acquisition of infection with Mycobacterium tuberculosis and the occurrence of clinical tuberculosis, if it develops, often precludes identification of the source of infection. In children, however, it is logical to assume that because of the small number of potential sources and the limited period of time (the child's life time) in which infection could have taken place, a source could be identified relatively easily. This is not the instance for children in California with either tuberculosis or tuberculous infection. A source case was found for only slightly more than one third of children younger than 5 yr of age reported with tuberculosis in 1994. In San Francisco, children who had a positive tuberculin skin test and no evidence of active tuberculosis, less than 2% had a likely source identified (1). A potential reason for the failure to identify sources of tuberculous infection in children is that the infection may have been acquired outside of the area in which the child resides, especially if the child travels to areas where there is a relatively high prevalence of tuberculosis. There is little information about the relationship between travel and acquisition of tuberculous infection, but one study found an increased risk of tuberculosis among adult Asians in London who had traveled to high-prevalence countries (2). Other studies, using molecular epidemiologic techniques, have identified tuberculosis strains indigenous to a particular area or country in adults who developed tuberculosis after travel to that area (3).

We conducted a case-control study to assess whether there is increased risk of tuberculous infection in children who have traveled to or been in a household with visitors from countries that have a high prevalence of tuberculosis. As noted, young children are an ideal study population because tuberculous infection in children indicates recent exposure to an infectious source. Previously unidentified risk factors for tuberculous infection in this population could shed light on the epidemiology of tuberculosis in the United States.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

Study participants were seen at public health clinics that largely serve populations covered by Medi-Cal insurance (California Medicare) or are eligible (family income at twice the federal poverty level) for the Child Health and Disability Prevention Program. A convenience sample consisting of 953 children in California younger than 6 yr of age had tuberculin skin tests performed between March and November 1996. The children had skin tests because they were at risk for infection with M. tuberculosis by living in areas that had a high incidence of tuberculosis, they or their parents were born in countries other than the United States, a skin test was required for school entry, or they had a known exposure to a person who had infectious tuberculosis (7). The children lived in seven counties of California, 57% in four northern California counties and 43% in three southern California counties. The incidence of tuberculosis in the counties where the children lived ranged from 12 to 36 cases per 100,000 population (weighted average, 26): the state average was 14 cases per 100,000 population in 1995 (8).

Study Design

The clinics were in California counties that had tuberculosis incidence rates approximate to or exceeding the state average for 1995. The standard practice at all participating clinics was to provide routine tuberculin skin tests as part of the child's annual health care maintenance. Tuberculin skin tests were applied using the Mantoux method, as recommended by the American Academy of Pediatrics (9), and read by a trained healthcare professional 48 to 72 h after application. All clinics used Tubersol (Connaught) for skin test antigen. A skin test was interpreted as positive based on the recommended stratification by risk factors (10). For most children a positive skin test resulted in induration of  10 mm and for contacts of tuberculosis cases induration of > 5 mm.

Questions about travel and household visitors in the preceding 12 mo were asked before the tuberculin skin test was read (either in English or in the primary language of the parent or guardian) to minimize recall bias. Travel was defined as a trip of more than 1 wk to a country with a high prevalence of tuberculosis. High-prevalence countries were estimated to have a tuberculosis rate of > 100 cases/100,000 and were in Latin America, Asia or the Pacific Ocean, and Africa (11). Questions about travel included the country visited, the length of stay, and the relationship of the person with whom the child stayed. The number of trips outside the United States that the child had taken was determined. Questions were also asked about household visitors from countries other than the United States who had stayed in the child's home for more than 1 wk. Specifically, parents or guardians were asked the country the visitor was from, the length of stay, the visitor's relationship to the child, and whether the visitor had any illness or symptoms suggestive of tuberculosis. In addition, demographic information (the date and place of birth, sex, race, and ethnicity) and information about bacille Calmette Guérin (BCG) vaccination and any previous skin test results was obtained. Physical examination was not performed to search for a visible BCG vaccination scar because such scars may be variably present (12, 13) and when present may not necessarily be associated with a delayed sensitivity response to skin testing (12).

Statistical Analysis

Demographic and clinical characteristics were compared among children who had a positive or negative tuberculin skin test using the chi-square test (Yates-corrected statistic or, when appropriate, Fisher's exact test) for categorical variables, the z score more normally distributed continuous variables, and the Mann-Whitney test for nonparametric data. In a univariate analysis we used unadjusted odds ratios (Mantel-Haenszel's technique) and 95% confidence intervals to assess whether travel to or having visitors from countries that have a high tuberculosis prevalence increase the risk of tuberculous infection in young children. After the analysis was done on the study population as a whole, the analysis was repeated on children who were born in the United States, and on children who were never vaccinated with BCG. In this way, existing cases from exposures that might be attributed to birth in a country that has a high tuberculosis prevalence or cases from boosting of reactions caused by previous BCG vaccination were eliminated. A secondary analysis of only children not known to have an exposure to a tuberculosis case was also done. The proportion of positive skin tests attributable to travel or to a household visitor was calculated using the following formula: attributable proportion = (test rate of the exposed group  test rate of the unexposed group)/ test rate of the exposed group. Multiple logistic regression analysis was used to examine the independent effects of each variable on the tuberculin skin test result after controlling for possible confounding variables (16). For all statistical tests, differences were considered significant at p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Characteristics of Children with a Positive Test

Seventy-two of the 953 children (7.6%) had a positive tuberculin skin test (Table 1); median induration was 15 mm. The mean age of children who had a positive skin test was 46 mo and 42 mo for children who had a negative skin test. Children who had a positive tuberculin skin test were more often female and had foreign-born parents than were children who had a negative skin test. Additionally, they were less likely to have been born in the United States or have had a previous skin test.

Although most children (84.3%) were born in the United States, 85.5% had a parent from a country other than the United States. About half (51.1%) of all children had at least one negative tuberculin skin test before the current test. The 120 foreign-born children who had an arrival date in the United States were in this country for a median of 18.4 mo (range, 0 to 60 mo) before the skin test. Of the 134 children (14.3%) who had BCG vaccination, the median time from vaccination to the current skin test was 46 mo (range, 6 to 72 mo) which did not differ significantly between children who had a positive or a negative test (p = 0.87). Only one of the five children who had been vaccinated within the 12 mo before the test had a positive reaction.

Travel to a High-Prevalence Country

Travel was common in the study population. In the 12 mo preceding the tuberculin skin test, 105 children (11%) had traveled to one of 15 high-prevalence countries. More than half (55%) of the children stayed with grandparents during their travel. Excluding day trips at the United States-Mexico border, 211 children (22%) had traveled at some time (mean lifetime travel was 1.6 trips per child). No child had traveled more than once in the previous year and none of the 32 African-American children had traveled. The median and mean lengths of travel were 1 and 6 mo (range, 0.3 to 48 mo). The countries traveled to reflect the demographics of this population: the areas most commonly visited were Mexico (64%); China, Hong Kong, and Taiwan (21%); and Central America (95%). Travel was more common among Asian children (17%), children 36 to 47 mo of age (16%), and children born in the United States (12%).

Of the 953 children studied, those who had traveled recently to a high-prevalence country were 4.9 times more likely to have a positive skin test than were children who had not traveled (22.9 versus 5.7%; 95% confidence interval [CI], 2.8 to 8.8; p < 0.001) (Table 2). After excluding 58 children who had had a household visitor, the proportion of positive skin tests attributable to travel was 77.6%. There of the four children who had a known tuberculosis exposure abroad had a positive test. When the seven children who had a known exposure to a tuberculosis case in the United States were excluded from the analysis, the odds ratio (OR) for a positive skin test was not changed appreciably (OR = 5.1; p < 0.001).

Children who had traveled and were either born in the United States or who had not had BCG vaccination were more likely to have a positive skin test (Table 2). After elimination of 17 children who had both traveled and had a household visitor, travel to a country having a high prevalence of tuberculosis remained associated with a positive test (OR = 4.9; 95% CI, 2.6 to 9.0; p < 0.001). Foreign-born children who had traveled did not have a significant increase in the proportion who had a positive test compared with those who had not traveled (OR = 2.8; 95% CI, 0.7 to 10.9; p = 0.14).

Household Visitors from a High-Prevalence Country

Fifty-eight of the 925 children (6.3%) for whom information was available had a household visitor in the 12 mo before their tuberculin skin test. The mean stay for visitors was 4 mo (range, 0.2 to 24 mo). Visitors were most commonly from Mexico (64%); China, Hong Kong, and Taiwan (16%); and Central America (11%). A visitor was more common in the household of children not born in the United States (10%) than those born in the United States (6%). Asian and Hispanic children had the same frequency of visitors (7%), and visitation did not vary significantly by age. Only one child had visitors on more than one occasion in the previous 12 mo.

Children who lived in a household that recently had a visitor from a country that has a high tuberculosis prevalence were 3.1 times more likely to have a positive skin test than were children who did not have a visitor (19 versus 7%; 95% CI, 1.4 to 6.6; p = 0.003) (Table 3). Children born in the United States and children who had never received BCG vaccination were also more likely to have a positive test if they lived in a household that recently had a visitor from a high-prevalence country (OR = 3.4 and 3.3, respectively). Foreign-born children who had a household visitor did not have a significant increase in the proportion who had a positive test (OR = 1.8; 95% CI, 0.5 to 6.4; p = 0.33).

Factors Associated with a Positive Tuberculin Test

After control for possible confounding covariates, travel to and a household visitor from a high-prevalence country remained significantly associated with a positive skin test (OR = 3.9 and 2.4, respectively) (Table 4). Travel was also associated with a positive test in the subpopulations of children who were born in the United States and those who had never had a BCG vaccination. Other significant factors for a positive test were female sex in all three populations and BCG vaccination in children born in the United States. The length of travel had a high degree of colinearity with travel itself; thus, we could not determine whether the length of travel was significantly associated with a positive test.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The principal findings of this study are that both travel to a country having a high prevalence of tuberculosis and contact with household visitors from such countries increase the risk of infection with M. tuberculosis in young children. Although travel has been suggested to put people at risk for tuberculosis, this study provides evidence that not only travel exposures but also household visitors from countries having a high tuberculosis prevalence present risks for tuberculosis infection. Evidence for these findings is provided by the logistic regression analysis, which showed an increased risk for a positive tuberculin skin test. Also, the proportion of positive tests attributable to travel indicates that travel is an important factor in the acquisition of tuberculous infection in young children. The evidence is strengthened by studying a population of young children, in whom the infection is more likely to have been recently acquired.

The infection rates in this study are unlikely to have been overestimated by falsely positive skin tests from mycobacteria-other-than-tuberculosis or BCG vaccination. Mycobacteria-other-than-tuberculosis usually do not produce a positive tuberculin skin test reaction (17); furthermore, infection from mycobacteria-other-than-tuberculosis is not common among children in Mexico (18), the country most frequently visited. Although recent BCG vaccination can cause induration of more than 10 mm, the vaccinated children had received BCG a median of 46 mo before the test, a period long enough after vaccination that most delayed-hypersensitivity reactions from BCG vaccination could be expected to wane (12, 14, 19). In a recent study among children 3 to 60 mo of age in Botswana mostly BCG vaccinated at birth, only 7% of the 820 children tested had a positive tuberculin reaction (22). The one factor in that study associated with a positive skin test was contact with an active case of tuberculosis, indicating that the tuberculin skin test is useful in diagnosing tuberculous infection in a population immunized with BCG.

Any selection bias that may have been present when we enlisted clinics for the study is unlikely to have affected the results. The children were seen at clinics that routinely test children for tuberculous infection. Another possible selection bias is that not all children tested returned for the skin test reading. For example, at one clinic 62% of children tested returned to have the skin test read. Children who returned did not differ from those who did not return in their age, sex, race, or ethnicity. The observed rate of return was comparable to previous reports, ranging from 47 to 65% (23). We believe that the study population was representative of young children who are tested for tuberculous infection in public health clinics in California, and inferences made from the data in this study apply to similar populations. Recall bias may have been present if the parents of children with a positive skin test were more likely to remember possible exposures. We attempted to minimize recall bias by concentrating the study on events from the 12 mo preceding the skin test.

Our findings have two important public health implications. First, tuberculosis control programs should ask about travel and household visitors in their investigations of children who had a positive tuberculin test. If the investigation of family and household members of a child who had a positive skin test does not identify a source case, but does find that the child has a recent history of travel to a country having a high tuberculosis prevalence, tuberculosis control programs should consider not extending contact investigations beyond the household.

Second, these findings have public health implications for screening populations at risk for tuberculosis. Children who have a history of travel to or a visitor from a high-prevalence country should be considered at increased risk for tuberculosis and these children should be tested. Targeted screening of children at risk for tuberculosis is cost-effective (24) and allows for prophylaxis using isoniazid at an age when efficacy approaches 100% (25, 26). In this study of children in areas of California known to have a high prevalence of tuberculosis, we found a tuberculin reactivity rate (7.6%) higher than the average for children screened for tuberculous infection on entering public school kindergarten in several areas of California (1.5%) (24, 27, 28). These rates for a positive test are substantially higher than other areas of the country for children entering kindergarten (27, 29).

When implementing screening for tuberculosis, the limitations of this study should be considered. It is possible that the study findings may not be generalizable to areas of the United States that have a low prevalence of tuberculosis. For the most part, we studied children who were born of immigrant parents, a situation that contributed to the frequency of travel by these children, and were of lower socioeconomic status. The study findings also need verification among foreign-born children who represented only 16% of the study population. Presumably, travel did not increase the risk for a positive skin test among the foreign-born because their primary exposure was prior to immigrating. Second, although the children in our study who had a positive test had characteristics similar to California children younger than 6 yr of age reported in 1995 as having tuberculosis (8), our study population underrepresented African-American children, who may have other risk factors for tuberculosis. Finally, other factors that play a role in the acquisition of tuberculosis, including household crowding (30), exposure to adults with risk factors for tuberculosis (31), underlying immunodeficiencies (32), and adult immigration status (33), were not addressed in this study.

In conclusion, we recommend that children who have traveled to or had visitors from a country having a high tuberculosis prevalence should be screened for tuberculosis. It is reasonable to skin test children after extended travel or frequent short visits to countries where tuberculosis is prevalent. In addition, when tuberculosis control programs investigate children who have tuberculous infection or disease, these programs should include questions about travel and household visitors in an attempt to identify the source of the child's infection. Our findings underscore the importance of defining risk factors for tuberculous infection and of using these risk factors to target populations for screening and to identify situations in which source cases are unlikely to be found.