Rationale: The goal for tuberculosis (TB) elimination in the United States is a TB disease incidence of less than 1 per million U.S. population by 2010, which requires that the latent TB infection (LTBI) prevalence be less than 1% and decreasing.
Objectives: To estimate the prevalence of LTBI in the U.S. population.
Methods and Measurements: Interviews and medical examinations, including tuberculin skin testing (TST), of 7,386 individuals were conducted in 1999–2000 as part of the National Health and Nutrition Examination Survey (NHANES), a nationally representative sample of the civilian, noninstitutionalized U.S. population. LTBI was defined as a TST measurement of ≥10 mm. Associations of age, race/ethnicity, sex, poverty, and birthplace were assessed. Results among the 24- to 74-year-old subgroup were compared with NHANES 1971–1972 data.
Measurements and Main Results: Estimated LTBI prevalence was 4.2%; an estimated 11,213,000 individuals had LTBI. Among 25- to 74-year-olds, prevalence decreased from 14.3% in 1971–1972 to 5.7% in 1999–2000. Higher prevalences were seen in the foreign born (18.7%), non-Hispanic blacks/African Americans (7.0%), Mexican Americans (9.4%), and individuals living in poverty (6.1%). A total of 63% of LTBI was among the foreign born. Among the U.S. born, after adjusting for confounding factors, LTBI was associated with non-Hispanic African-American race/ethnicity, Mexican American ethnicity, and poverty. A total of 25.5% of persons with LTBI had been previously diagnosed as having LTBI or TB, and only 13.2% had been prescribed treatment.
Conclusions: In addition to basic TB control measures, elimination strategies should include targeted evaluation and treatment of individuals in high-prevalence groups, as well as enhanced support for global TB prevention and control.
Estimated national prevalence of latent tuberculosis infection in the United States decreased from 14.3% in the early 1970s to 5.7% in 1999–2000 among persons 24–74 years of age. Rates of active TB disease also decreased from 15.8 to 5.6 per 100,000 during the same period in the U.S. population as a whole, but both of these reported decreases mask continuing high rates among important groups such as the foreign-born population and U.S.-born ethnic minorities.
Based on a representative national survey, latent TB infection prevalence in the United States was 4.2% in 1999–2000, with significantly higher rates in some subgroups. Only 25.5% of persons with latent TB infection had been diagnosed, and only 13.2% had been prescribed treatment.
Current guidelines for implementing the IOM recommendations include targeted tuberculin skin testing (TST) for groups at higher risk for TB, including individuals born outside the United States. Nationally, there are no published current estimates on the percentages of individuals in these groups who have had TST, have LTBI, or have been treated for LTBI. In addition, all groups that should be targeted may not have been identified. Because the risk of progression from LTBI to TB is life-long, estimates of differential risk in U.S. population subgroups based only on TB case reports could underestimate the risk for groups in whom TB is underdiagnosed. As TB rates decline, sources of continued transmission will increasingly be those not identified by current surveillance methods (5).
Estimates of LTBI, testing, and treatment in high-risk subgroups are needed. To achieve TB elimination, Styblo calculated that the LTBI rate must be less than 1% and decreasing (6). Large, representative, population-based LTBI surveys with sufficient numbers to represent subgroups are difficult, and none were performed in the United States for decades after the first National Health and Nutrition Examination Survey (NHANES), which included an LTBI substudy in 1971–1972. This report describes the findings of the NHANES 1999–2000 TB component, which included an LTBI assessment. Some preliminary analyses from this study were previously reported in the form of abstracts (7, 8).
NHANES is a series of cross-sectional, nationally representative health examination surveys (9). Each survey provides national prevalence estimates, enabling examination of trends over time. A complex, stratified, multistage probability cluster sampling design is used to select a nationally representative sample of the U.S. civilian, noninstitutionalized population. The current NHANES, which began in 1999, produces an annual national probability sample. NHANES 1999–2000 oversampled low-income persons, adolescents aged 12 to 19 years, persons aged 60 years and older, African Americans, and Mexican Americans to support separate analyses in these groups.
This report is based on NHANES 1999–2000 data for 7,386 participants with TST results. For the subgroup aged between 25 and 74 years, results were compared with NHANES 1971–1972, which included a TB component for that age group.
Self-reported race and ethnicity were categorized as “non-Hispanic black,” “non-Hispanic white,” and “Mexican American.” Persons not fitting these categories were classified as “other” and included in the population totals. Age was defined as the age at the time of the household interview, which preceded TST examinations by 2–6 weeks.
The poverty:income ratio was calculated by dividing the family income (all income received in the previous 12 mo) by the poverty threshold, adjusted for family size for the interview year, as determined by the U.S. Bureau of the Census. Poverty was defined as a poverty:income ratio below 1.
TST was performed using purified protein derivative (PPD) S-1 (PPD-S), the national standard tuberculin developed by Seibert, against which U.S. commercial tuberculin preparations are tested (10). To measure potential cross-reactivity to nontuberculous mycobacteria (NTM), PPD-Battey (PPD-B), prepared from the Boone strain of Mycobacterium intracellulare and shown to be the most useful NTM antigen for detecting infection by common NTM (11), was also used; both antigens were used in NHANES 1971–1972. Because neither is commercially licensed for clinical use, an informed consent procedure was used under Investigational New Drug Protocol 7596 approved by the Food and Drug Administration.
For each participant, a randomization computer program selected which PPD should be applied to each forearm. Trained NHANES phlebotomists injected 0.1 ml of the designated PPD intradermally to the volar surface of the designated forearm using the Mantoux method. Participants were asked to return at 48–72 hours after TST placement for measurement of reactions by trained NHANES TST readers, who had no knowledge of which PPD had been applied to either arm or of participants' TB-related history.
Reactions to both PPDs were recorded for each participant by at least one trained reader who measured the extent of induration in mm. Study protocol dictated that two separate readers blinded to one another's measurements would measure TST reactions of 25% or more of the participants. Readers worked in separate rooms and did not have access to one another's measurements, which were recorded in a computer database. For the analysis, the mean TST reaction and PPD-B reaction were used for each participant who had reactions read by two or more readers.
The definition of LTBI was an induration of 10 mm or greater in reaction to TST. An alternate LTBI definition was also used: a TST reaction 15 mm or greater or a TST reaction of between 5 and 15 mm, and at least 2 mm greater than the reaction to PPD-B. The use of PPD-B to inform interpretation of TST results is derived from a 1960s study of Navy recruits diagnosed with LTBI (12), indicating that individuals with TST reactions between 6 and 12 mm, which were at least 2 mm greater than their reactions to PPD-B, had a similarly high risk of developing TB disease as individuals with a TST reaction of greater than 12 mm, while those with TST reactions between 6 and 12 mm that were smaller than their reactions to PPD-B had as low a TB risk as those with TST reactions of 5 mm or less. Because true LTBI prevalence was likely to be much lower in the 1999–2000 U.S. population than in the Navy study, we conservatively required TST reactions 13–14 mm, as well as those 6–12 mm, to be 2 mm greater than the PPD reaction to meet the LTBI alternate definition. Because the United States' cut-off defining LTBI requiring treatment is 15 mm for persons without identified risks (13), a result 15 mm or greater defined LTBI in the alternate definition, regardless of PPD-B result.
The household interview included six TB questions, the answers to which were used to define participants' TB history. These questions are listed in the online supplement.
NHANES medical examiners were trained to recognize bacillus Calmette-Guérin (BCG) scars, and to differentiate them from smallpox vaccination scars through supervised examination of photographs and human volunteers. Visible scars evaluated by examiners as BCG scars were recorded.
Data management, variable definitions and recodes, and preliminary analyses were performed in SAS (SAS Institute, Cary, NC). Percent prevalence of TST positivity, numbers with LTBI, and respective 95% confidence intervals (CIs) were calculated using PROC CROSSTAB in SUDAAN, version 9.0 (RTI International, Research Triangle Park, NC). Factors associated with LTBI were examined using multivariate logistic regression (RLOGIST in SAS-callable SUDAAN).
Prevalence estimates for the U.S. population were calculated using NHANES 1999–2000 Medical Examination Center sample 2-year weights (to adjust for unequal probability of selection and nonresponse to the household interview and physical examination) (14); additional weights were calculated for TST nonparticipation (15). Details on weighting and treatment of missing values are in the online supplement.
Univariate associations between LTBI and demographic factors were examined by the Cochran-Mantel-Haenszel test for unequal odds. Multivariate analyses were performed using logistic regression. The main outcome variable in our model was a TST reaction of 10 mm or greater. The predictor variables were sex, age group, race/ethnicity, foreign birth, and poverty:income ratio. The effect of age was assessed using a test for trends. The original model included all first-order interactions between predictor variables. We selected 0.05 as the level of statistical significance. The weighted odds ratios (ORs) and 95% CIs were based on standard errors incorporating the complex sample design. A detailed discussion of the calculation of sample variances, which differs from the methodology used in NHANES 1971–1972, can be found on the NHANES website (9).
To examine changes over time, data from NHANES 1971–1972 were reanalyzed using the 1971–1972 NHANES Medical Examination Center weights, adding an additional weight for nonparticipation in TST, as was done for the 1999–2000 dataset, and applying the same basic definition of LTBI.
Computer-generated reports were read aloud to and given to each participant, with instructions to seek additional medical evaluation for a “positive” TST. Participants received contact information for local public health department TB control programs, the staffs of which were contacted in advance by NHANES staff to alert them that referred individuals might seek evaluation. Because of NHANES confidentiality restrictions, NHANES staff could not contact TB control programs directly about individuals, and it was not possible to retrieve follow-up information about individuals to whom follow up was recommended. No referrals were made for PPD-B results.
In 1999–2000, 8,832 individuals eligible for TST placement had the NHANES medical examination. Of these, 7,613 (86%) had a TST placed; of those with a TST, 7,386 (97%) had their reactions measured. In NHANES 1971–1972, 1,494 adults aged 25–74 years had TST reactions measured.
Based on our weighted analysis, we estimate 4.2% of the civilian, noninstitutionalized U.S. population aged 1 year or older, or 11,213,000 individuals, had LTBI in 1999–2000. Based on reported history, we estimate that 25.5% (16.6–37.1%) of those with LTBI had previously been diagnosed with LTBI or TB, and 13.2% (8.6–19.7%) of those with LTBI had previously been treated for LTBI or TB.
Cross-sectional 1999–2000 prevalence estimates and estimated numbers with LTBI, with 95% CIs for major subgroups, are presented in Table 1.
LTBI Prevalence % (95% CI)
Population with Characteristic, n (× 1,000)
No. (× 1,000) with LTBI (95% CI)
|All participants||4.2 (3.3–5.2)||268,284||11,213 (8,938–14,038)|
|Male||5.2 (4.2–6.4)||131,052||6,747 (5,455–8,324)|
|Female||3.2 (2.3–4.6)||137,232||4,434 (3,110–6,294)|
|Age group, yr|
|1–14||1.1 (0.5–2.4)*||57,262||636 (296–1,357)|
|15–24||2.4 (1.1–5.0)*||39,352||942 (449–1,948)|
|25–44||5.0 (3.7–6.6)||84,806||4,227 (3,170–5,612)|
|45–64||6.5 (4.5–9.5)||56,455||3,681 (2,511–5,341)|
|⩾65||5.6 (3.7–8.4)||30,409||1,713 (1,135–2,562)|
|Non-Hispanic white||1.9 (1.3–2.9)||182,745||3,548 (2,361–5,313)|
|Non-Hispanic black/African American||7.0 (5.3–9.1)||32,036||2,244 (1,711–2,928)|
|Mexican/Mexican American||9.4 (7.6–11.5)||20,815||1,958 (1,589–2,402)|
|Other||10.8 (7.2–15.9)||32,688||3,523 (2,341–5,198)|
|Poverty income index ⩾1||3.3 (2.5–4.4)||15,496||6,488 (4,859–8,637)|
|Poverty income index <1||6.1 (4.0–9.1)||45,810||2,791 (1,853–4,160)|
|<High school||5.6 (4.4–7.1)||100,292||5,523 (4,368–6,961)|
|High school graduate||3.4 (2.3–5.0)||52,679||1,804 (1,232–2,630)|
|Beyond high school||3.7 (2.4–5.7)||97,314||3,628 (2,373–5,507)|
|U.S.||1.8 (1.4–2.1)||231,227||4,154 (3,073–5,607)|
| Foreign||18.7 (13.5–25.2)||37,057||6,888 (4,993–9,292)|
Prevalence of LTBI among males was 5.2% compared with 3.2% among females. LTBI prevalence increased significantly with age group between the 1- to 4-year-old group (1.1%) and the 45- to 64-year-old group (6.5%,); among the ⩾65-year-old age group, LTBI prevalence was 5.6%. Non-Hispanic whites had the lowest LTBI prevalence (1.9%); a significantly higher prevalence was seen both among non-Hispanic blacks (7.0%) and Mexicans/Mexican Americans (9.4%).
LTBI prevalence was significantly higher among those living in poverty (6.1%) compared with those living above the poverty level (3.3%), and significantly higher for those with less than a high school education (5.6%) compared with high school graduates (3.4%) and those with education beyond high school (3.7%).
Foreign-born participants had a significantly higher LTBI prevalence (18.7%) than participants born in the United States (1.8%).
Because of the marked difference between the LTBI prevalences in the United States–born and foreign-born groups, separate analyses were performed for these two groups. Cross-sectional prevalence estimates of LTBI for the United States–born and foreign-born populations are presented separately in Table 2. Among the foreign-born population, each demographic subgroup for which comparisons were made had a significantly higher prevalence of LTBI than the comparable subgroup born in the United States.
|LTBI Prevalence||Population with Characteristic||Estimated No. with LTBI||LTBI Prevalence||Population with Characteristic||Estimated No. with LTBI|
|Characteristics||% (95% CI)||No. × 1,000||No. × 1,000 (95% CI)||% (95% CI)||No. × 1,000||No. × 1,000 (95% CI)|
|All participants||1.8 (1.3–2.4)||231,227||4,154 (3,073–5,607)||18.7 (13.5–25.2)||37,057||6,888 (4,993–9,292)|
|Female||1.5 (1.0–2.2)||112,019||1,761 (1,202– 2,575)||14.4 (8.7–23.0)||19,033||2,596 (1,565–4,138)|
|Male||2.1 (184.108.40.206)||119,208||2,380 (1,482–3,802)||22.7 (16.3–30.5)||18,023||4,313 (3,109 5,812)|
|Age group, yr|
|1–14||0.3 (0.1–1.1)*||53,781||173 (49–610)||11.9 (5.2–24.8)*||3,480||413 (182–861)|
|15–24||0.6 (0.2–1.6)*||33,597||193 (68–540)||12.8 (5.1–28.4)*||5,755||735 (295–1,636)|
|25–44||1.2 (0.7–2.2)||68,841||826 (453–1,500)||20.6 (14.6–28.1)||15,965||3,281 (2,330–4,492)|
|45–64||3.4 (2.0–5.8)||48,189||1,650 (963–2,797)||25.3 (17.6–35.1)||8,266||2,094 (1,454–2,897)|
|⩾65||4.8 (2.8–8.0)||26,819||1,288 (762–2,148)||11.9 (5.2–24.8)*||3,590||427 (188–899)|
|Non-Hispanic white||1.1 (0.6–2.0)||17,441||1,960 (1,120–3,418)||17.9 (11.4–26.8)||8,333||1,489 (953–2,236)|
|Non-Hispanic black/African American||5.7 (4.2–7.8)||29,193||1,661 (1,212–2,263)||20.0 (13.7–28.4)||2,842||570 (388–808)|
|Mexican/Mexican American||2.5 (1.6–3.8)||12,372||307 (200–470)||19.1 (16.2–22.5)||8,443||1,616 (1,366–1,900)|
|Other||1.5 (0.4–5.2)*||15,249||221 (60–794)||18.6 (10.9–29.9)||17,438||3,241 (1,900–5,211)|
|Poverty income index|
|Poverty income index ⩾1||1.4 (1.0–2.1)||171,561||2,469 (1,714–3,550)||16.5 (11.8–22.7)||23,936||3,950 (2,813–5,428)|
|Poverty income index <1||2.8 (1.9–4.0)||3,751||1,052 (728–1,516)||20.3 (13.0–30.3)||8,295||1,687 (1,082–2,514)|
|<High school||2.5 (1.7–3.5)||81,900||2,003 (1,391–2,874)||19.2 (14.9–24.4)*||18,392||3,527 (2,734–4,483)|
|High school graduate||1.6 (0.9–2.8)||46,842||740 (413–1,316)||17.9 (8.9–32.7)||5,837||1,046 (520–1,911)|
| Beyond high school||1.6 (1.0–2.7)||85,094||1,371 (824–2,273)||18.3 (9.5–32.1)||12,220||2,230 (1,164–3,923)|
Similar differences in prevalence between males and females were seen both among the United States–born (2.1 vs. 1.5%) and the foreign-born (22.7 vs. 14.4%) groups. A significantly higher prevalence was observed among those living in poverty compared with those living above the poverty line in both the United States–born (2.8 vs. 1.4%) and the foreign-born (20.3 vs. 16.5%) groups.
For other subgroups, different patterns appeared among the two populations. Among the those born in the United States, non-Hispanic blacks and Mexicans/Mexican Americans, respectively, had significantly higher prevalences than non-Hispanic whites (5.7 and 2.5 vs. 1.1%). Among the foreign born, prevalences did not differ significantly among racial/ethnic groups.
Separate logistic regression analyses were performed for factors associated with the outcome of LTBI using poverty:income ratio and education level, respectively, as the surrogate for socioeconomic status. Both analyses produced similar results; the model with poverty:income ratio produced narrower CIs and are presented here.
For the U.S. population as a whole, logistic regression with a model including all categories in the univariate analyses and all first-order interactions showed the following variables as independent positive predictors of LTBI: non-Hispanic black race/ethnicity; Mexican/Mexican American ethnicity; age; poverty; and foreign birth. Interactions were observed between foreign birth and race/ethnicity and between foreign birth and age group (data not shown).
In a separate model including only participants born in the United States (see Table 3), the following were significant independent predictors of LTBI: male sex (adjusted OR, 1.9); poverty (OR, 1.9); non-Hispanic black race/ethnicity (OR, 7.5); Mexican-American ethnicity (OR, 5.2); age 15–24 years (OR, 2.0); age 25–44 years (OR, 6.0); age 45–64 years (OR, 21.4); age 65 years or older (OR, 34.3); and poverty (OR, 1.9) (test for trend, P < 0.01).
|Factor||Adjusted OR (95% CI)||Adjusted OR (95% CI)|
|Non-Hispanic black||7.5 (4.0–13.9)||1.0 (0.5–1.8)|
|Mexican American||5.2 (2.7–10.0)||0.9 (0.5–1.6)|
|Other||2.6 (0.9–7.8)||1.0 (0.5–2.0)|
|Poverty income ratio ⩾1||1.0||1.0|
|Poverty Income ratio <1||1.9 (1.3–2.9)||1.7 (0.9–3.3)|
|Male||1.9 (1.0–3.4)||2.0 (1.0–4.0)|
|Age group, yr|
|15–24||2.2 (0.2–20.8)||1.0 (0.3–4.1)|
|25–44||6.0 (2.9–12.1)||2.0 (0.9–4.7)|
|45–64||21.4 (8.4–54.5)||3.0 (1.1–8.2)|
| ⩾65||34.3 (6.4–185)||1.0 (0.4–2.4)|
In the model with foreign-born participants (see Table 3), significant independent predictors of LTBI were male sex (OR, 2.0) and age 45–64 years (OR, 3.0).
In 1999–2000, estimates based on self-report suggest that 67.9% of the U.S. population had had a previous TST; of those, 5.1% had a history of a positive result. Of all the U.S. population, both with and without a previous TST, 3.5% had been told previously that they had LTBI or TB, and 1.3% had been prescribed treatment. Details on TB-related history can be found in Table E4 in the online supplement.
Of those with LTBI, 35.9% of the United States born versus 19.5% of the foreign born with LTBI had a previous diagnosis of LTBI or TB, and 15.8% of the United States–born versus 11.6% of the foreign-born groups had had treatment for LTBI or TB.
Based on reported history, we estimate that, of the 1999–2000 U.S. population who had lived with a person with TB, 87.2% had had a TST. Of those, 16.6% (10.5–25.2%) had a history of a positive TST result, and of those, 50.7% had a history of LTBI treatment. A total of 2.3% (0.8–6.2%) had been previously diagnosed with TB; of those, 98.8% (88.4–99.9%) had been prescribed treatment.
Based on NHANES examiners' reports, an estimated 25.0% (16.1–36.7%) of foreign-born individuals had BCG scars. A total of 86.8% (72.2–93.8%) of the foreign-born individuals with TST of 10 mm or greater, with a BCG scar identified, were 25 years old or older. If foreign-born individuals identified with a BCG scar who did not report a history of TB or LTBI were considered not to have LTBI, the estimated LTBI prevalence would be 3.5% (2.7–4.3%) in the U.S. population and 16.4% (12.3–21.4%) in the foreign-born U.S. population. These estimates do not differ significantly from the estimates reported without considering BCG scars.
The distribution of TST reactions from 0 to 32 mm estimated for the 1999–2000 U.S. population is shown in Figure 1; 82.1% of TST reactions were 0, and 11.5% of reactions were between 1 and 4 mm. In the reactions greater than 5 mm, the mode (0.66%) is seen at 10 mm.
Using the alternate definition of LTBI that incorporated a comparison of the TST and the PPD-B reactions, the 1999–2000 prevalence of LTBI would be estimated at 3.8% (3.1–4.7%) rather than 4.2% in the total U.S. population, 1.6 (1.2–2.2%) versus 1.8% in the United States–born population, and 17.3 (13.5–21.9%) versus 18.7% in the foreign-born population. The estimated number of infected individuals would be 10,269,000 (3,723,000 United States born and 6,386,000 foreign born). No differences between the respective estimates are significant.
With the alternate definition incorporating the PPD-B reaction measurement, a bimodal distribution of the reactions of 5 mm or greater was observed, with modes at 10 mm (0.39%) and 15 mm (0.39%) (data not shown).
Among the subgroup 25–74 years of age, we estimate that LTBI prevalence decreased from 14.3% (11.3–18.0%) in 1971–1972 to 5.7% (4.5–7.2%) in 1999–2000, a decrease of 60% (P < 0.05). This corresponds to LTBI in 14,671,000 individuals in 1971–1972 and 9,069,000 in 1999–2000, a decrease of 38% in absolute numbers.
This report provides the first survey-based national LTBI estimates since 1971–1972. Between 1971–1972 and 1999–2000, LTBI rates among adults 25–74 year old decreased by 60%, from 14.3 to 5.7%. In comparison, TB case rates per 100,000 population reported in the U.S. TB surveillance system decreased by 63%, from 15.8 to 5.6, between 1972 and 2000 (2). As seen in TB case reports, the decrease in the United States LTBI rate masks the higher rates seen among Mexican Americans and African Americans born in the United States, in persons born outside the United States, and in persons living in poverty. The LTBI rates among non-Hispanic whites, 1.9%, is close to that required for TB elimination, but the far higher rates among all other groups, and the relatively small number of individuals this group contributes to active TB cases (2), make TB elimination by 2010 in the United States unlikely.
In 2002, for the first time, reported TB cases among foreign-born persons constituted the majority (51%) of cases in the United States (16). In 1999–2000, we estimate that 63% of TB infections in the United States were among foreign-born individuals. Only 19.5% of the foreign born with LTBI had been previously diagnosed with LTBI or TB, and only 11.6% had ever been prescribed treatment, as compared with 35.9 and 15.8% of the United States–born population with LTBI. This left an estimated pool of 6,148,000 (5,479,000–6,536,000) foreign-born and 3,542,000 (3,154,000–3,806,000) United States–born individuals with LTBI with a potential risk of disease progression, which would vary based on factors not evaluated in NHANES, including immune status and recentness of infection (17). Although some who were diagnosed and not prescribed treatment may have been correctly evaluated as unlikely to progress to TB disease, which would lower these estimates, others who were prescribed treatment may not have begun or completed it, which would raise the estimates. Studies in the United States indicate that the percentage of persons who complete treatment for LTBI may be 65% or higher in some areas, but may be 51% or lower in others (18–22).
LTBI rates have increased among the foreign born between 1971–1972 and 1999–2000, although the estimate for 1971–1972 is less stable, given the smaller numbers surveyed. According to U.S. Census data, there were 9.6 million foreign-born individuals in the United States in 1970 (23), making up only 4.7% of the population; this had increased to 31.1 million (10.4%) by the year 2000 (24). Much of the increase in population is associated with immigration from countries with a high prevalence of TB. In 1970, the leading countries of origin for foreign-born U.S. residents were six European countries and Mexico, Canada, the Soviet Union, and Cuba (25); in 2000, the leading countries of origin included Mexico, China, the Philippines, India, Vietnam, Cuba, Korea, Canada, El Salvador, and Germany (24). All the leading countries of origin in the Year 2000, except Cuba, Canada, and Germany, were among the 22 countries with the highest burden of TB worldwide in the early 2000s (26). Although information on country of origin is not available for NHANES 1999–2000, it is likely that much of the foreign-born increase in LTBI prevalence is associated with the changes in the makeup of the foreign-born population in the United States. These findings justify enhancing efforts to support global TB control and a more systematic evaluation of LTBI and appropriate treatment and follow-up services to immigrants, as recommended by the IOM (5). The finding that approximately one-third of the foreign-born U.S. population had no previous TST to screen for TB infection, despite the fact that this is a high-risk group, supports the IOM's recommendation for improved outreach to the foreign-born population in the United States.
It is possible that some foreign-born persons defined as having LTBI might have had a TST reaction of 10 mm or greater resulting from BCG vaccination rather than LTBI. However, BCG-induced TST reactions have been observed to wane over time, and are considered negligible after 10–15 years (27, 28). Of the foreign-born group with TST of 10 mm or greater who were identified as having BCG scars, 87.5% were 25 years old or older, and unlikely to have received BCG within the past 10 years, given that BCG is usually administered in infancy or early adolescence. In practice, the clinical definition for LTBI requiring preventive treatment for the foreign born in the United States is a TST reaction of 10 mm or greater, regardless of BCG history (29, 30). Because of the high TB prevalence of TB in most countries with routine BCG use, the likelihood that a positive TST indicates TB infection outweighs the likelihood of a false-positive result caused by BCG. Many persons identified with a BCG scar in NHANES reported a history of TB, LTBI, or TB-related treatment.
United States–born, non-Hispanic whites are the only group surveyed with an LTBI prevalence (1.1%) close to the less than 1% that is required for TB elimination. Among those born in the United States, higher LTBI rates were seen among non-Hispanic blacks and Mexican Americans compared with non-Hispanic whites. These findings are consistent with the TB case reports for the Year 2000, in which non-Hispanic blacks had a TB rate of 12.3 per 100,000 population, more than twice that of the United States–born Hispanic rate of 5.2, which was nearly twice the United States–born non-Hispanic white rate of 2.8 (2). Previous analyses have suggested that socioeconomic status, and specifically poverty, may account for many of the apparent racial discrepancies seen in TB rates in the United States (31), but race/ethnicity remained a predictor of LTBI in United States–born individuals even after adjustment for poverty. Among the foreign born, all three major ethnic groups had equivalent LTBI rates, substantially higher than those of any United States–born group.
The finding that poverty remained a predictor of LTBI in the United States–born population after adjustment for confounding by sex, race/ethnicity, and age is important. TB case reporting does not include socioeconomic status, and targeting for evaluation and treatment has focused on groups known to be at high risk through TB case reports. The fact that homeless and incarcerated individuals, who make up much of the population living in poverty in the United States and who have high rates of TB (32–36), were not included in NHANES makes the association between LTBI and poverty especially striking. Although TB diagnosis and treatment are free through public health departments in most of the United States, outreach to individuals living in poverty, including those who are neither homeless nor incarcerated, should increase.
This survey has several limitations. The use of the poverty:income ratio as the sole indicator of poverty is a potential limitation. The exclusion of the homeless and the incarcerated may have resulted in a lower LTBI prevalence estimate than would be found in a survey including these groups. The fact that Asians, Asian Americans, Native Americans, and Pacific Islanders could not be oversampled to provide separate estimates is unfortunate, because TB rates are high in these groups, but their LTBI risks could not be separately assessed. Given the geographic variability in TB case rates in the United States, another limitation is that we could not analyze geographic LTBI variations, because of potential instability of estimates for this low-prevalence condition based on relatively low numbers surveyed in each region of the United States. Performance of the NHANES TB component of the survey for a longer time period, when it is next implemented, could provide higher numbers for area-based estimates.
Estimating proportions of individuals with a history of TST, LTBI, TB, and treatment based on self-report may have lead to bias. Individuals may not remember a previous test, or may have misidentified a TST as an immunization. Reports of a previous positive TST may be overestimated, as individuals may have mistakenly reported a remembered erythematous reaction or a small induration as “positive.” Reports of previous TB may be overestimated if individuals previously diagnosed with LTBI confused TB infection with disease. Reports of prescribed medicine may be underestimated by those who did not actually fill prescriptions or complete treatment, but a regimen lasting at least 6 months is likely to have been correctly reported by those who experienced it, so the history of treatment for LTBI and TB are likely to represent a reasonable, minimal estimates.
TST remains an imperfect tool for diagnosis of LTBI, and, as the prevalence of LTBI decreases, the positive predictive value of TST decreases. Also, false-positive TST reactions are associated with many causes, including cross-reactions from NTM infections, nonstandardized application of antigen, TST reader inexperience, and variations in reading techniques. However, in the NHANES TB component, training in standard methods of application and reading, supervision, and comparisons of blinded dual readings were included to standardize procedures and to increase and assess reliability. The similar distributions of TST measurements, and the similar prevalences estimated using the reactions recorded by different readers, suggest that reliability was reasonable. The higher prevalence of a previous history of LTBI or TB (25.5%) among individuals with LTBI compared with that in individuals without LTBI (2.5%) provides an additional rough validation of the NHANES TST measurement methods. Last, the alternate definition incorporating a comparison of the reaction to an NTM antigen to the tuberculin reaction arrived at similar prevalence estimates to those using the standard definition, indicating that cross-reactivity may not have substantially biased the estimate upward.
The fact that 77.4% with LTBI in the U.S. population are estimated to have had a previous TST, but that only 33.0% had history of a positive result, does not mean that either previous TST readings or current NHANES readings were inaccurate. TB infection could have occurred subsequent to the historical measurements in many individuals with positive NHANES TST reactions. The fact that household TB contacts reporting previous TST had similar estimated LTBI rates based on NHANES and reported historical results (15.2 and 14.8%, respectively) indicates a high degree of consistency in TST readings for this group, whose historical TST were likely to have occurred soon after infection.
Despite the use of PPD-B, some false-positive results are likely to have been included in the estimate, given the low prevalence of LTBI in the United States. However, the fact that, above 10 mm, there were relatively few TST reactions for which the PPD-B reaction was 2 mm greater indicates that cross-reactivity with NTM may not have been a major source of overestimation. Reversion, in which individuals with previously positive TST demonstrate a subsequent negative reaction, could have led to a counterbalancing underestimation of LTBI in some individuals. Anergy, which is seen in some immunocompromised individuals, is unlikely to have been a major source of bias, given the low prevalence of HIV and other immunocompromising conditions in the U.S. population, but no estimate of the association of immune deficiency with TST results is possible with available NHANES data.
The use of the 10-mm cutoff, although it makes this survey comparable to others performed internationally, may have biased the estimate upward. The current U.S. cutoff for defining LTBI requiring preventive treatment is 15 mm when no risks (e.g., HIV infection or recent close contact with TB) are ascertained to warrant use of a lower cutoff. Most risk information was not ascertainable in NHANES; some individuals with a TST of 10 mm or greater but less than 15 mm may not have had LTBI that would be likely to progress to active TB or to require preventative treatment under current guidelines. On the other hand, some with TST of greater than 5 mm and less than 10 mm might have met the clinical criteria for preventive treatment with a clinical evaluation. The 10 mm mode for TST results greater than 5 mm suggests that the use of 10 mm as the cut-off for LTBI may be reasonable. Also, the use of the alternate definition for LTBI, which incorporated PPD-B to evaluate TST results between 5 and 14 mm, showed modes at both 10 and 15 mm, and did not result in a significantly different LTBI prevalence estimate. Unfortunately, analytic techniques (37, 38) to increase sensitivity and specificity of LTBI prevalence estimates cannot easily be applied to NHANES where individual results are differentially weighted to represent different population numbers. We will explore the applicability of these techniques in the future.
The lack of a chest radiograph in NHANES 1999–2000, and the confidentiality restrictions that made follow up impossible, means that, theoretically, individuals with active TB could have been included in the LTBI estimate. However, active TB in NHANES 1971–1972, as identified by the NHANES chest radiograph, was so rare that a positive TST could be considered as equivalent to LTBI. Given that there was a significantly lower prevalence of TB in the United States in 1999–2000 than in 1971–1972 (16), it is likely that a positive TST is also equivalent to LTBI in NHANES 1999–2000.
Despite the limitations, NHANES 1999–2000 provides important information on the prevalence of LTBI in the civilian, noninstitutionalized U.S. population, and in many important groups. Oversampling in Mexican Americans, adolescents, the elderly, and the poor supported stable estimates for these important groups that would not have been otherwise available. The 4.2% rate of infection in the population overall makes TB elimination unlikely by 2010, and the substantially higher LTBI rates among some subgroups suggest that TB elimination will require specific public health actions for these groups. This analysis reinforces the IOM's conclusion that, in addition to continuing basic TB control through treatment for LTBI and TB, TB elimination strategies should include targeted evaluation and appropriate treatment of individuals in high-prevalence groups. In addition to greater efforts to screen and appropriately treat immigrants, support for global TB prevention and control efforts in high-burden countries is, as Bloom noted (39), not only a matter of humanitarian concern, but a matter of enlightened self-interest. Our findings suggest that global efforts could contribute directly to TB elimination in the United States. Finally, better access to TB services for and improved outreach to people living in poverty are needed to further TB elimination efforts in the United States.
The following individuals contributed substantially to the planning or analysis of this research: Jose Becerra, Ph.D., and George Cauthen, Ph.D. (retired), U.S. Centers for Disease Control and Prevention, Atlanta, Georgia; Andrew Margileth, M.D., Memorial Health University Physicians, Savannah, Georgia; Michael J. Brennan, Ph.D., U.S. Food and Drug Administration, Rockville, Maryland.
|1.||Cantwell MF, Snider DE, Cauthen GM, Onorato IM. Epidemiology of tuberculosis in the United States, 1985 through 1992. JAMA 1994;272:535–539.|
|2.||U.S. Centers for Disease Control and Prevention. Reported tuberculosis in the United States, 2005 [Internet]. Atlanta (GA): U.S. Department of Health and Human Services; 2006 Sep [accessed 2007 Jan 7]. Available from:|
|3.||U.S. Centers for Disease Control and Prevention. A strategic plan for the elimination of tuberculosis in the United States. MMWR Morb Mortal Wkly Rep 1989;38:1–25.|
|4.||Geiter L, editor. Ending neglect: the elimination of tuberculosis in the United States. 1st ed. Washington (DC): National Academy Press; 2000.|
|5.||Onorato I. TB outbreaks in the United States. Int J Tuberc Lung Dis 2000;4:S121–S126.|
|6.||Styblo K. The elimination of tuberculosis in the Netherlands. Bull Int Union Tuberc Lung Dis 1990;65:49–55.|
|7.||Bennett DE, Onorato IM, Courval J, Agerton T, Daugherty-Gibson J, Lambert L, McQuillan G, Lewis B, Castro K. Prevalence of TB infection in the U.S. population [abstract 385]. 33rd International Union of Tuberculosis and Lung Disease Conference on Lung Health, Paris, France; November 2002.|
|8.||Bennett DE, Onorato IM, Courval J, Agerton T, Daugherty-Gibson J, Lambert L, McQuillan G, Lewis B, Castro K. Prevalence of TB infection in the U.S. population [abstract 67921]. 131st American Public Health Association Annual Meeting, San Francisco, California; November 2003|
|9.||U.S. Centers for Disease Control and Prevention. NHANES analytic guidelines. June 2004 version [accessed 2007 Aug 1]. Available from:|
|10.||Seibert FB, Munday B. Chemical composition of active principle of tuberculin; precipitated purified protein suitable for preparation of standard tuberculin. Am Rev Tuberc 1934;30:482–484.|
|11.||Rust P, Edwards LB, Hopwood L, Palmer CE. Identification of mycobacterial infections. Bull World Health Organ 1965;33:405–412.|
|12.||Edwards LB, Acquaviva FA, Livesay VT. Identification of tuberculous infected: dual tests and density of reaction. Am Rev Respir Dis 1973;108:1334–1339.|
|13.||ATS/CDC Statement Committee on Latent Tuberculosis Infection. Targeted tuberculin testing and treatment of latent tuberculosis infection. MMWR Morb Mortal Wkly Rep 2000;49:1–51.|
|14.||U.S. Centers for Disease Control and Prevention. NHANES 1999–2000 data release sample weights to use with particular data files [accessed 2007 Jan 5]. Available from:.|
|15.||Kalton G, Flores-Cervantes I. Weighting methods. J Off Stat 2003;19:81–97.|
|16.||U.S. Centers for Disease Control and Prevention. Trends in tuberculosis morbidity—United States, 1992–2002. MMWR Morb Mortal Wkly Rep 2003;51:217–224.|
|17.||Horsburgh CR Jr. Priorities for treatment of latent tuberculosis infection in the United States. N Engl J Med 2004;350:2060–2067.|
|18.||Marks S, Taylor Z, Shrestha-Kuwahara R, Wilce M, Nguyen C. Outcomes of contact investigations of infectious tuberculosis patients. Am J Respir Crit Care Med 2000;162:2033–2038.|
|19.||Reichler MR, Reves R, Bur S, Ford J, Thompson V, Mangura B, Onorato IM, Valway SE; Contact Investigation Study Group. Treatment of latent tuberculosis infection in contacts of new tuberculosis cases in the United States. South Med J 2002;95:414–420.|
|20.||Tulsky JP, Pilote L, Hahn JA, Zolopa AJ, Burke M, Chesney M, Moss AR. Adherence to isoniazid prophylaxis in the homeless: a randomized controlled trial. Arch Intern Med 2000;160:697–702.|
|21.||LoBue PA, Moser KS. Use of isoniazid for latent tuberculosis in a public health clinic. Am J Respir Crit Care Med 2003;168:443–447.|
|22.||Cook PP, Maldonado RA, Yarnell CT, Holbert D. Safety and completion rate of short-course therapy for treatment of latent tuberculosis infection. Clin Infect Dis 2006;43:271–275.|
|23.||Gibson CJ, Lennon E. Nativity of the population and place of birth of the native population: 1850 to 1990 [Internet]. Washington (DC): U.S. Census Bureau; 1999 [revised 2001; accessed 2007 Jan 7]. Available from:|
|24.||Malone N, Baluja K, Costanzo J, Davis C. The foreign born population 2000. Census 2000 brief [Internet]. Washington (DC): U.S. Census Bureau; 2003 [accessed 2007 Jan 7]. Available from:|
|25.||Gibson CJ, Lennon E. Region and country or area of birth of the foreign-born population: 1960 to 1990 [Internet]. Washington (DC): U.S. Census Bureau; 1999 [revised 2001; accessed 2007 Jan 7]. Available from:|
|26.||World Health Organization. 2003 Global tuberculosis control: surveillance, planning, financing. Geneva: World Health Organization; 2003. WHO/CDS/TB/2003.316.|
|27.||U.S. Centers for Disease Control and Prevention. The role of BCG vaccine in the prevention and control of tuberculosis in the United States. MMWR Recomm Rep 1996;45:1–18.|
|28.||Wang L, Turner MO, Elwood RK, Schulzer M, FitzGerald JM. A meta-analysis of the effect of Bacille Calmette Guérin vaccination on tuberculin skin test measurements. Thorax 2002;57:804–809.|
|29.||Advisory Council for the Elimination of Tuberculosis. Essential components of a tuberculosis prevention and control program: recommendations of the Advisory Council for the Elimination of Tuberculosis. MMWR Recomm Rep 1995;44:1–16.|
|30.||U.S. Centers for Disease Control and Prevention. Recommendations for prevention and control of tuberculosis among foreign-born persons. MMWR Recomm Rep 1998;47:1–29.|
|31.||McKenna MT, McCray E, Jones JL, Onorato IM, Castro KG. The fall after the rise: tuberculosis in the United States, 1991 through 1994. Am J Public Health 1998;88:1059–1063.|
|32.||Kong PM, Tapy J, Calixto P, Burman WJ, Reves RR, Yang Z, Cave MD. Skin-test screening and tuberculosis transmission among the homeless. Emerg Infect Dis 2002;8:1280–1284.|
|33.||Haddad MH, Wilson TW, Ijaz K, Marks FM, Moore M. Tuberculosis and homelessness in the United States, 1994–2003. JAMA 2005;293:2762–2766.|
|34.||Stead WW. Undetected tuberculosis in prison: source of infection for the community at large. JAMA 1978;240:2544–2547.|
|35.||Spencer SS, Morton AR. Tuberculosis surveillance in a state prison system. Am J Public Health 1989;79:307–310.|
|36.||MacNeil JR, Lobato MN, Moore M. An unanswered health disparity: tuberculosis among correctional inmates, 1993 through 2003. Am J Public Health 2005;95:1800–1805.|
|37.||Reider HL. Methodological issues in the estimation of the tuberculosis problem from tuberculin surveys. Tuber Lung Dis 1995;76:114–121.|
|38.||Neuenschwander B. Bayesian mixture analysis for tuberculin induration data for the International Union against Tuberculosis and Lung Disease, July 2007 [accessed 2007 Aug 10]. Available from:|
|39.||Bloom BR. Enlightened self-interest and the control of tuberculosis. N Engl J Med 2005;353:1057–1058.|