Published ahead of print on February 8, 2008, doi:10.1164/rccm.200711-1756OC
© 2008 American Thoracic Society doi: 10.1164/rccm.200711-1756OC
Evaluation of a Model for Efficient Screening of Tuberculosis Contact Subjects 1 Service de Pédiatrie, and 2 Centre de Lutte Anti-Tuberculeuse, Centre Hospitalier Intercommunal de Créteil, Créteil, France; 3 Association Clinique et Thérapeutique Infantile du Val de Marne, Saint Maur des Fossés, France; 4 INSERM U550, Faculté Necker-Enfants Malades, Paris 5, Paris, France; 5 Institut de Veille Sanitaire, Département des Maladies Infectieuses, Saint Maurice, France; 6 INSERM, U841, Créteil, France; and 7 Université Paris 12, Faculté de Médecine, IFR10, Créteil, France Correspondence and requests for reprints should be addressed to Christophe Delacourt, M.D., Ph.D., Service de Pédiatrie, Centre Hospitalier Intercommunal de Créteil, 40 avenue de Verdun, 94000 Créteil, France. E-mail: christophe.delacourt{at}chicreteil.fr
Rationale: Contact tracing is an important component of tuberculosis (TB) control programs. Standardization of contact investigation protocols can make them more efficient. Objectives: To develop a model to select contact subjects for screening. Methods: We prospectively collected standardized data on 325 TB index cases and their 2,009 contacts. Factors that independently influenced the risk of TB infection were included in the model, which was then validated in a second prospective cohort of 88 cases of TB and their 618 contacts.
Measurements and Main Results: A total of eight independent risk factors were identified (odds ratio; 95% confidence interval): age, with three subgroups: 6–14 years (3.6; 1.6–8.0); 15–29 years (3.7; 1.8–7.7); Conclusions: This study provides a standardized contact screening model which reduces resources required without negatively affecting disease control.
Key Words: tuberculosis • contact screening • tuberculin skin test
Tuberculosis (TB) is one of the leading causes of disease and death worldwide (1). In France, most cases are diagnosed in the Paris region, where the incidence is four times the national average (2). TB control programs must include efficient detection and treatment of latent infection. Screening for TB can be improved by using a standardized, validated questionnaire to identify contacts who qualify for tuberculin skin testing (TST). The use of such models can significantly reduce the number of contacts who need to be screened, while maintaining good disease control. Validated models for TB contact screening are limited. In a low-incidence setting, a model with seven variables significantly predicted positive TST results among contacts of a patient with active TB (3). The model had a sensitivity, specificity, and positive predictive value of 89, 36, and 26%, respectively. However, TST results were considered positive when the induration measured 5 mm or more, and this cutoff may not be appropriate for countries such as France that have low to moderate incidence rates and high bacillus Calmette-Guérin (BCG) vaccination coverage (4). Furthermore, important risk factors, such as closeness of nocturnal contacts, were not included in this model (5). To establish a screening model for populations with high BCG vaccination rates, we prospectively identified risk factors for TB infection among subjects exposed to a contagious case using a standardized data collection process. The resulting model was then validated in a second sample of consecutive cases of TB and their contacts.
Study Population The study participants lived in the Val de Marne suburb of Paris, where the incidence of TB, deduced from the reported rate, was 22.1 per 100,000 inhabitants in 2005 (2). We enrolled consecutive subjects exposed to a patient with pulmonary TB between April 2004 and December 2005. An index case was defined as the first individual in whom TB was diagnosed. All culture-positive patients older than 15 years of age with newly diagnosed pulmonary TB were included in the study. We also included proven pediatric cases when screening was undertaken in the institutions they attended. A contact subject was defined as any subject living in the environment of a TB case during the 3 months preceding diagnosis of the index case (6). Contact subjects with a documented history of a TST induration of 15 mm or more were excluded. To validate the model, we performed the same analysis on a second sample consisting of 88 consecutive cases of TB and their 618 contacts enrolled between January and December 2006 in the Val de Marne region of Paris.
Screening Organization Contact subjects are investigated according to national guidelines published by Société de Pneumologie de Langue Française (7) and Conseil Supérieur d'Hygiène Publique de France (8). For this study, the first screening visit (V1) included a physical examination, TST, and chest radiograph. These investigations were repeated 8–12 weeks later (V2) if the contact subject did not meet the criteria for infection at V1. When the interval between the last exposure to the index case and V1 exceeded 8 weeks, no further investigations were undertaken. All contacts who were under 2 years of age or who had underlying health problems were offered chemoprophylaxis, whatever their initial results. Otherwise, chemoprophylaxis was proposed to contact subjects meeting the diagnostic criteria for latent infection. Contacts with abnormal chest radiographs were further evaluated for TB disease and were treated if disease was confirmed.
Diagnostic Criteria for Infection in Contact Subjects In symptom-free contacts with normal chest radiographs, latent TB infection (LTBI) was diagnosed on the basis of the TST performed on the volar surface of the forearm by intradermal injection of 0.1 ml (5 TU) of purified protein derivative tuberculin obtained from a human Mycobacterium tuberculosis strain (Tubertest; Sanofi Pasteur, Lyons, France). The test was read within 72 hours by measuring diameter of the induration. Contact subjects were classified as infected if the induration was 10 mm or more in the absence of prior BCG vaccination. BCG-vaccinated contacts were classified as infected if the induration was 15 mm or more at V1 or V2, or if it converted from negative at V1 (<5 mm) to 10 mm or more at V2 (7, 8). These cutoffs were previously demonstrated to identify infected contacts, despite BCG vaccination (4, 9–12).
Risk Factors Index case infectivity was assessed from the duration of cough before TB diagnosis, cavitation on chest radiography, extent on chest radiography, and bacillary density in sputum smears and culture, graded with international semiquantitative scales (13). HIV serostatus and homeless status were not included in the analysis of risk factors for contact infection because of the small number of contacts with these characteristics.
Statistical Analysis Application of the model requires the selection of a predictive probability above which all contacts must be examined. To determine this cutoff point, we calculated the sensitivity, specificity, positive predictive value, and negative predictive value of different probability levels. Receiver operating characteristic curves were constructed for both the model dataset and the validation dataset, plotting true-positive cases (sensitivity) and false-positive cases (1 – specificity) for each probability level. Data were analyzed using StatView 5.0 software (SAS Institute, Cary, NC), and were expressed as mean values (±SEM).
Study Population During the study period, 325 index cases were included. The index cases had a mean age of 41.5 (±1.1) years (range, 3–96 yr). Cough was present at diagnosis in 89% of cases. The mean duration of cough before diagnosis was 2.6 (±0.2) months. Cavitations were found on the chest radiograph in 41% of index cases. Sputum smears were positive for acid-fast bacilli in 59% of index cases. A total of 2,009 contacts were identified. A total of 54 contact subjects who had a history of a TST induration of 15 mm or more were excluded. The characteristics of the index cases and their screened contacts are shown in Table 1.
The mean number of contacts per case was six (range, 1–122). The screened contacts had a mean age of 28.5 (±0.3) years (range, 3 d–85 yr). Nearly all the contacts (98%) had been vaccinated with BCG. Data on BCG vaccination were extracted from documented medical records (45%), from identification of a BCG scar (37%), and from interview (16%). Among contacts with documented medical records, the mean delay between last BCG vaccination and screening was 12.8 (±0.3) years. The first screening visit (V1) took place a mean of 43 (±0.8) days after diagnosis of the index case. The screening results are shown in Figure 1. The 380 contacts with incomplete screening were excluded from the final analysis. Their mean age, sex, relationship to the index case, and mean duration of contact did not differ significantly from the corresponding characteristics of contacts with complete screening. The second screening visit (V2) took place a mean of 82 (±1) day after V1. The overall infection rate among the 1,575 contacts with complete screening was 27.0%. A total of 15 contacts (1%) had active TB and 410 (26.0%) had latent infection. Among the 366 children under 15 years of age, 48 (13.1%) had latent infection and 5 (1.4%) had overt disease.
Risk Factors for Contact Infection Univariate analysis. We examined potential risk factors for transmission among the characteristics of the contacts, index case infectivity, and exposure. The risk of TB infection among contacts was significantly related to markers of socioeconomic status and individual risk factors, such as smoking (Table 2). The risk of a positive TST leading to diagnosis of TB infection also increased significantly with age and with the incidence of TB in the contact's country of birth. This suggested that a significant proportion of TB infections identified in adult contacts were not due to recent exposure.
The influence of index case infectivity on the risk of contact infection is shown in Table 3. The risk of TB infection among contacts was significantly increased when the index case was a smoker, had cavitation on the chest radiograph, or was sputum smear positive. Moreover, the risk of infection increased gradually with the density of acid-fast bacilli in the index patient's sputum.
Finally, the risk of contact infection was strongly associated with the time spent with the index case (Table 4). Mean number of hours spent with the index case was 321 (±22) hours in infected contacts, compared with 211 (±11) hours in uninfected contacts (P < 0.0001); this increased significantly with close contact at night, and with the time spent with the index case during the day. The risk of contact infection increased with the familial relationship between the contact and the index case: first-degree relatives were significantly more likely to be infected than were more remote family members and contacts unrelated to the index case.
Multivariate analysis. Multivariate analysis identified eight independent factors significantly associated with the risk of being infected (Table 5): age; cavitation on the index case's chest radiograph; an index case sputum smear with 100 or more acid-fast bacilli per field; household contact at night; first-degree family relationship with the index case; active smoking by the contact; free health care; and birth in a country with a TB incidence rate higher than 25/100,000. Our contact screening model included these eight risk factors. A hypothetical contact subject with all eight risk factors had a 0.92 probability of being infected, whereas a contact subject with none of the eight risk factors had a 0.033 probability of being infected (see Table E6 in the online supplement). To ensure that our results were not biased by the choice of an arbitrary cutoff for TST, we performed a step-by-step regression analysis with test reading as a dependent continuous variable, and with the eight factors incorporated in our model as independent variables (1 = factor not present; 2 = factor present). All 8 variables were retained in this regression model, attesting to the robustness of our model, independently of the chosen cutoff point. Finally, because several contacts came from the same index case, and to ensure that a cluster effect did not bias our multivariate analysis, a general estimating calculation was performed with the eight parameters included in our model. Coefficients were found to be at the same level, attesting to the validity of the interactions that we found between risk factors.
We then calculated the sensitivity, specificity, positive predictive value, and negative predictive value for each level of predictive probability (Table E7). In order that the targeted contact screening would not readily miss recently infected contacts, we chose a probability level with which the false-negative rate was close to the background rate of preexisting infection. To estimate this background in our study population, we analyzed contact subjects with prior TST. A prior TST result was available for 569 contacts, and was 15 mm or more in 54 subjects. The rate of preexisting TB infection in the whole population was therefore estimated to be 9.6%. However, age and birth country significantly influenced the risk of having a previous TST with induration greater than 15 mm; adjusted odds ratios (95% confidence intervals) were 2.39 (1.02–5.60) for ages 15–30 years, 6.33 (2.40–16.72) for over 30 years of age, and 3.15 (1.71–5.82) for those born in a country with an incidence higher than 25 of 100,000 inhabitants. The background LTBI is therefore dependent on these two factors. The widest range is age related: 3.9% (6/153) in children younger than 15 years of age, 9.2% (31/338) in 15- to 30-year-old contacts, and 21.8% (17/78) in contacts older than 30 years of age. To ensure picking up the right group for intervention, we chose probability cutoffs yielding a false-negative rate lower than the estimated background prevalence of latent infection. We chose a predictive probability of 0.08 for contacts younger than 15 years of age, 0.15 for contacts of 15–30 years of age, and 0.20 for contacts older than 30 years of age. Corresponding false-negative rates were 0, 9.1, and 15.5%, respectively. The use of these adjusted predictive probabilities reduced the number of contacts to be investigated by 26% while maintaining an overall false-negative rate of 8%. Sensitivity, specificity, and the positive predictive value of this model with adjusted cutoffs were 0.93, 0.34, and 0.36, respectively. Finally, we used these adjusted probabilities to test our screening model in a new prospective sample of 88 index cases and their 618 contacts. In this new sample, the sensitivity, specificity, positive predictive value, and negative predictive value of the model were 0.94, 0.32, 0.42, and 0.91, respectively. The false-negative rates remained lower than the expected background in the different age groups: 3.2, 1.9, and 20.3% in contacts younger than 15 years of age, 15–30 years of age, and older than 30 years of age, respectively. Only one child under 15 years of age, with a 17-mm TST reading and a normal chest X-ray, would have been missed by our screening model. The false-negative rate was 9% for the whole population. The percentage of unscreened persons was 20.3%. Receiver operating characteristic curves for the model sample and the validation sample were very similar (Figure 2).
Standardization improves the efficiency of contact investigations (14). We therefore prioritized data helpful both for contact investigation and treatment decisions, and validated a screening model based on eight factors that independently influenced the risk of infection through contact with TB disease.
Accurate diagnosis of LTBI in contacts depends largely on TST. However, in countries with extensive BCG vaccine coverage, BCG vaccination may affect the interpretation of skin test results. A tuberculin reaction of 15 mm or more in diameter is generally considered to be a good criterion for identifying latent M. tuberculosis infection in vaccinated persons. This position came from both longitudinal and cross-sectional studies. In a longitudinal study, Chee and colleagues (9) examined the risk of TB disease in BCG-vaccinated children for the 4 years subsequent to TST reading at age 12 years. They found that a TST reading of 15 mm or more predicted the development of TB disease with a specificity of 98.2%, and a sensitivity of 27.1% (9). In a cross-sectional study, Menzies and colleagues found that the percentage of schoolchildren with TST readings of 15 mm or more varied from 1 to 3% according to socioeconomic status (12). Finally, a recent meta-analysis confirmed that patients who had received BCG vaccination were more likely to have a positive skin test, and that indurations larger than 15 mm were more likely to result from TB infection than from BCG vaccination (4). Sensitivity of this 15 mm cutoff, estimated in patients with TB disease, ranged from 75 (11) to 83% (10). With this 15-mm cutoff, the proportion of infected contacts was very similar to that found in the United States, where about 20–30% of all contacts have latent infection and 1% have TB disease (15). With a 10-mm cutoff in our population, 59% of contacts would have been considered infected (data not shown). Similarly, in a study of a large group of contacts of several active cases of TB, ignoring the BCG history led to significant overuse of chemoprophylaxis. Indeed, 18% of 732 unvaccinated contacts were offered chemoprophylaxis, compared with 44% of 386 contacts who had received BCG (16). IFN- Although many factors significantly influence the risk of TB infection, there are few data on their interactions (3, 19, 20). The total duration of exposure is a well-recognized risk factor for TB infection (3, 5, 21). We found that sharing the same home at night was the best parameter for evaluating the duration of exposure. Furthermore, the closeness of contact was an additional risk factor, independent of the duration of exposure. Intrafamilial contact is also known to increase the risk of TB infection (3, 21). We found a direct relationship between genetic relatedness and the risk of TB infection. After adjustment for other risk factors, being a first-degree relative of an index case remained a significant predictor. This not only confirms the role of close contact with the index case in the transmission of TB infection but might also suggest genetic susceptibility to M. tuberculosis infection. The infectivity of the index case was a significant risk factor in many studies (3, 22, 23). Our results underline the importance of bacillary density in sputum smears, as only a high density of acid-fast bacilli was an independent risk factor for TB infection, in addition to cavitation on the chest radiograph. Of particular interest is the role of active smoking in TB infection shown here. Active and passive smoking is an identified risk factor (24, 25), but only one recent study identified active smoking as an independent risk factor for TB infection (26). Finally, we found that the risk of TB infection increased with age, especially after 15 years of age. This relationship was still significant when age was adjusted for the TB incidence in the birth country. Repeated prior contact with M. tuberculosis in the birth country is the most likely explanation for this finding. Indeed, in our initial population, age and country of birth were found to significantly influence the risk of having a previous TST with induration greater than 15 mm. This failure to differentiate recent infection from background infection is not specific to TST, but is also shared by IGRAs. In situations with high community exposure to M. tuberculosis, enzyme-linked immunospot assays, whole-blood assays, and TSTs are each positive in a majority of healthy adult control subjects (27, 28). The 8% false-negative rate in our screening model corresponds to the fraction of contacts with TST indurations of 15 mm or more who were not investigated because their probability of being infected was below the adjusted cutoffs. Ideally, the false-negative rate should be lower or equivalent to the background rate of positive TST reactions. This background rate is not related to recent TB transmission, and ranges from 5 to 10% in countries with incidence rates similar to that in France (29, 30). It was shown to be dependent of age and birth country in our study population, and ranging from 3.9 to 21.8%. By adjusting the choice of probability to these background estimations, our model is very unlikely to miss many recently infected subjects with positive TSTs. Of particular interest, none of the children under 15 years of age in our validation sample, and only one in our confirmatory sample, who had a predictive probability below 0.08 and who would not, therefore, have been investigated in our screening model, was considered as infected. In conclusion, this study provides a basis for standardizing contact-tracing protocols and for a contact screening model that reduces resources devoted to screening without negatively affecting disease control. As such, this model, in the hands of experienced TB field workers, should help to ensure optimal contact screening.
Supported by Programme Hospitalier de Recherche Clinique AOR 04 003, AP-HP; Institut de Veille Sanitaire; and Legs Poix.
* A complete list of members may be found before the beginning of the REFERENCES. This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Originally Published in Press as DOI: 10.1164/rccm.200711-1756OC on February 8, 2008 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Members of the CG94 Study Group: Saïd Aberrane, Marc Angebault, Fadi Antoun, Jacqueline Argast, Jean Baptiste Armengaud, Kinan Atassi, Brigitte Aubert, Anandadev Banerjee, Sylvie Baot, Guilène Barnaud, Laurence Bassinet, Dany Baud, Halim Bekri, Ghéricia Benkabou, Jean Michel Berthuin, Xavier Blanc, Marie Josée Boivin, Jean Pierre Bonniot, Danièle Bosc, Geneviève Bosongo, Michel Boucherat, Catherine Branger, Françoise Brun Vesinet, Isabelle Caby, Emmanuelle Cambau, Amélie Carrere, Jean Didier Cavallo, Anne Chace, Mireille Cheron, Bertrand Chevalier, Michel Chousterman, Sylvie Coudray, Gaëlle Cuzon, Annie Dardour, Marie-Françoise David, Thierry Debord, Marion Decobert, Lionel Deforges, Isabelle Delacroix, Jean François Delfraissy, Catherine Deschamps, Jean Paul Dommergues, Alain Dublanchet, Elisabeth Dussaix, Jean-Philippe Emond, Christelle Epaud, Lelia Escaut, Elisabeth Estrangin, Rolland Fabre, Bruno Fantin, Nicolas Fortineau, Michel Fournier, Claire Fuhrman, Chantal Gagliardone, Jean Louis Gaillard, Eliane Garrade, Véronique Garrigue, Marion Gory, François Guillot, Sylvie Guy, Paul Hadchouel, Béate Heym, Jean Paul Homasson, Patrick Honderlick, Bruno Housset, Patrick Imbert, Jean Claude Jannaud, Gilles Jebrak, Sophie Larrar, Frédérique Lartigue, Christine Lascols, Nathalie Launay, Nicole Leflour, Agnès Lefort, Jeanne Marie Le Glaunec, Sylvie Le Maho, Philippe Lesprit, Corinne Levy, Jean Pierre L'Huillier, Nathalie Lorinet, Françoise Madre Pichon, Bernard Maitre, Jeannie Claude Malahel, Hedi Mammeri, Gilles Mangiapan, Armelle Marceau, Giselle Marcolino, Valérie Matharan, Marie Pierre Menager, Brigitte Ménégol, Bettina Mesplee, Sylvie Meto, Liliana Mihaila, Catherine Minozzi, Isabelle Monnet, Paul Morin, Marie Muntrez, Marie Hélène Nicolas Chanoine, Patrice Nordmann, Latifa Noussaire, Valérie Nouyrigat, Sonia Nunez-Godoy, Christine Orzechowski, Georges Otterbein, Olivier Patey, Christian Perronne, Nicole Pillet, Laurence Poluzzi, Céline Pothier, Jean Dominique Poveda, Françoise Querrec, Aimée Raharijaona, Linda Reure, Elisabeth Rivaud, Valérie Robin, Serge Roden, Raymond Ruimy, Marie Saillour, Annie Sfez, Djamel Smaine, Alain Sobel, Véronique Soudier, Claude James Soussy, Marc Stern, Claudette Sy, Marie Claire Tcholakian, Elina Teicher, Evelyne Thiaffey, Sadia Tortorelli, Sabine Trombert, Maryline Vergne, Dominique Vignon, Philippe Vinceneux, Daniel Vittecoq, Catherine Weil Olivier, Mona Yazji-Zarudiansky, and Laurence Zunic. Received in original form November 28, 2007; accepted in final form February 4, 2008
Related articles in AJRCCM:
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
30 years (4.1; 2.0–8.5); cavitation on the index case's chest radiograph (1.6; 1.1–2.2); an index case sputum smear with 100 or more acid-fast bacilli per field (1.8; 1.2–2.8); household contact at night (2.1; 1.3–3.2); first-degree family relationship with the index case (2.1; 1.3–3.3); active smoking by the contact (1.6; 1.1–2.4); free health care (2.0; 1.2–3.2); and birth in a country with TB incidence rate higher than 25 of 100,000 (2.2; 1.5–3.2). Predictive probabilities were chosen to ensure false-negative rates lower than estimated TB infection background. The number of contacts to be investigated was reduced by 26% while maintaining a false-negative rate of 8%. 
release assays (IGRAs) demonstrated very similar specificity values to those obtained through TST readings in BCG-vaccinated persons when a 15-mm induration cutoff value is used (
This work is dedicated to the memory of B. Decludt. 
