INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the United States, persons with newly reported tuberculosis ("index cases") are interviewed to identify individuals to whom they may have transmitted Mycobacterium tuberculosis (1). Those persons ("contacts") so identified are then evaluated to determine if they have tuberculous infection or active tuberculosis. Management of contacts is guided by two principles. First, rates of tuberculosis are known to be high in contacts, presumably because they have been infected recently (2, 3). This incidence can be reduced by the use of isoniazid preventive therapy given to contacts with infection but not disease (4, 5). Second, the organisms causing disease and/or infection in the contact are assumed to be the same strain that caused disease in the index cases and, thus, be susceptible to the same antimicrobial agents (6).

Using DNA fingerprinting we have compared isolates of M. tuberculosis from pairs of index and contact cases to test the assumption that tuberculosis in a contact is caused by the same strain that caused the disease in the index cases. In the setting of epidemiologically suspected transmission, the isolation from two persons of M. tuberculosis having the same DNA fingerprint provides compelling evidence that transmission has occurred (7).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Contact Tracing

The sampling frame was San Francisco County, January 1, 1991 to December 31, 1996. Names of contacts were elicited for all persons in whom infectious (pulmonary and laryngeal) tuberculosis was diagnosed and all persons with tuberculosis younger than 15 yr of age. The Centers for Disease Control and Prevention (CDC) criteria were used for the definition of a case of tuberculosis (10). Cases that met predefined criteria for laboratory cross-contamination were excluded. These criteria were as follows: (1) only one culture from that patient was positive and the specimen had a negative microscopic examination, (2) another patient's isolate processed at the same laboratory within 1 mo was smear-positive and had an identical DNA fingerprint, and (3) the clinical course was not suggestive of active tuberculosis.

For cases treated by the Division of Tuberculosis Control, trained disease-control investigators (DCIs) conducted contact investigations according to standard methods (1). Contact investigations for patients receiving care elsewhere were carried out either by the treating physician or by DCIs. Persons who spent an estimated total of at least 40 to 100 h with the index cases in the 3 mo prior to diagnosis or during the infectious period were considered to be "close contacts" and those who shared the same front door with the index case were considered "household contacts." The initial contact investigation was more extensive if organisms were detected by sputum microscopy. For index cases with positive sputum smears, evaluation of extended family members and contacts at the school, place of work, or social setting was initiated immediately. On the basis of the prevalence of positive tuberculin skin test results in the initial circle of contacts it was determined whether to expand the investigation in concentric circles as per the stone-in-the-pond principle (11). For index cases having negative smears but positive cultures, the initial investigation focused on close and household contacts, and the investigation was extended only if there was a higher-than-expected prevalence of positive tuberculin skin test results in the inner circle. For culture-negative index cases, investigation was limited to close and household contacts. For index cases younger than 15 yr of age, investigation directed toward finding a possible source was carried out, regardless of the site of disease. During 1993, contact investigations were broadened, especially among homeless and U.S.-born index cases; specifically, the number of cases without contacts was reduced and the number of contacts investigated was increased.

During the DCI's interview, contacts were queried concerning symptoms of tuberculosis, a history of tuberculin testing, and whether or not there was a previous diagnosis of tuberculosis. Tuberculin skin testing using 5 IU purified protein derivative was performed and read 48 to 72 h later. If the tuberculin test was positive (> 5 mm induration) a chest radiograph was obtained. Depending on the clinical evaluation and the results of the chest radiograph, individuals were classified as having only tuberculous infection or were investigated further for active tuberculosis. Contacts with a suspicion of tuberculosis were usually referred to the tuberculosis clinic for further evaluation, and a smaller proportion was evaluated by their private physicians. Contacts with tuberculous infection were usually offered chemoprophylaxis with isoniazid, with the exception of contacts of drug-resistant cases who were managed according to the drug-susceptibility result of the index case (12). Those who were found to have active tuberculosis were treated with multidrug chemotherapy (13).

Molecular Epidemiologic Investigation

In order to select index and contact cases for comparison, contacts with tuberculosis, either at the time of investigation or at a later date, were linked by name with the registry of all cases of active tuberculosis reported during the study period in San Francisco. Routinely collected information was extracted and coded for blinded analysis by tuberculosis control personnel. Demographic, clinical, and drug susceptibility data were available from the records maintained by the San Francisco Division of Tuberculosis Control and added by patient code. If two individuals had each named the other as a contact, only one pair was kept, with the first case diagnosed considered to be the index case. Because children are generally not considered to be contagious, when the first case was a child (younger than 15 yr of age), the adult was considered the index case (14). For other pairs we have defined cases as index and contact cases since we could not assign a source case with certainty.

Comparison of fingerprints required that both index and contact cases have interpretable RFLP patterns. IS6110-based RFLP was performed using the international standard method (15), with digitized images stored using the Whole Band Analyzer Version 3.3 (BioImage, Ann Arbor, MI). These digitized images were compared visually, and when in doubt original images were retrieved and compared directly. Pairs were classified as having the same strain if (1) their IS6110 patterns had greater than five bands and were identical or differed only by the addition, loss, or shift of one band (16), or (2) their IS6110 patterns had five or fewer bands of identical sizes and also had matching polymorphic-GC-rich sequence (PGRS) based RFLP patterns (17). For PGRS comparisons, unscanned films were compared directly.

We examined risk factors for transmission according to features of (1) exposure, (2) index cases, and (3) contact cases. Exposure categories were defined above as close and/or household. For index cases we examined age, sex, ethnicity, country of birth, HIV status, sputum-smear status, and presence or absence of cavitary disease on chest radiograph. HIV status was recorded at the time of tuberculosis diagnosis as known HIV-positive, known HIV-negative, or HIV status unknown. For contact cases we examined the same variables, with the exception of sputum-smear and radiographic findings.

Patient identifiers were removed prior to analysis to protect confidentiality. Because an index case could be linked to more than one contact, the number of pairs does not exactly equal the number of index cases. For statistical analysis of exposure, the unit of measurement was contact pairs. For analysis of index cases the unit of measurement was index cases, for analysis of contacts the unit of measurement was contact cases. EpiInfo Version 6 (CDC, Atlanta, GA) was used to generate 95% confidence intervals of point estimates and risk ratios for comparisons of proportions. We report all p values by Fisher's exact test less than 0.01, recognizing that the statistical treatment of multiple comparisons is controversial (18).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

During the study period, 1,855 cases of tuberculosis were reported in San Francisco. There were 1,468 cases of pulmonary tuberculosis, no diagnosed cases of laryngeal tuberculosis, and 57 cases of pediatric tuberculosis. At least one contact was identified for 1,308 (85.6%) of these 1,525 cases (see Figure 1). These 1,308 index cases resulted in 11,211 contact investigations (median of four per index case). In the latter half of the study period (1994 to 1996), there was an increase in the proportion of cases with at least one contact identified (from 696/850 = 0.82 to 627/675 = 0.93, RR = 1.13, p < 0.001). The prevalence of infection and disease among contacts were greater in close or household contacts (see Table 1). Contacts had a median age of 35 yr, a prevalence of tuberculous infection of 35.4% (3,976/11,211), and prevalence of active tuberculosis of 0.96% (108/11,211). Compared with contacts of U.S.-born cases, contacts of foreign-born cases had a significantly higher prevalence of tuberculous infection (43.7 versus 24.5%, p < 0.001) and a significantly lower prevalence of tuberculosis (0.6 versus 1.6%, p < 0.001). Of the 11,102 contacts without tuberculosis at the time of investigation, 11 subsequently developed tuberculosis after a median interval of 218 d.

View larger version (21K): [in this window] [in a new window]   Figure 1.   Tuberculosis contact investigations, San Francisco 1991-1996. Asterisk indicates dual listings; that is, index and contact had each named the other.

In total, there were 94 pairs of index and contact case with active tuberculosis (83 at the time of investigation and 11 subsequently) 66 of which had positive cultures. Of the 28 contact cases that were culture-negative, 17 were pediatric cases; in fact, 17/20 contact cases in children were culture-negative. Fingerprints were available for 54/66 pairs of index and contact cases. These 54 pairs comprised 54 different contact cases and 38 different index cases, five of whom had more than one contact. Thirteen (34%) index cases had only contacts with different fingerprints, two (5%) had contacts with both the same and different fingerprints and 23 (61%) had only contacts with the same fingerprint. A different fingerprint was seen in contacts of 15 of 38 index cases (39%; 95% CI: 24 to 57%) or in 16 of 54 pairs of index-contact cases (30%; 95% CI: 18 to 44%). The proportion of unmatched fingerprints did not increase significantly after 1993 when more extensive investigation occurred (9/28 versus 7/26, RR = 1.19, 95% CI: 0.52 to 2.74).

For the 38 pairs where the same fingerprint was found, drug susceptibilities for index and contact cases were the same. However, of the 16 pairs having different fingerprints, seven (44%) had different drug-susceptibility patterns (p < 0.001). The overall concordance of drug susceptibility between index and contact cases was 47/54 (87%). The proportion of matched fingerprints according to presence or absence of variables of interest is shown in Table 2. Unmatched fingerprints were statistically associated with two variables. Foreign-born contact cases were more likely to have different strains than U.S.-born contact cases (68.8 versus 13.2%, RR = 5.22, p < 0.001). Asian contact cases were more likely to have different strains than non-Asian contact cases (77.8 versus 20.0%, RR = 3.89, p = 0.002). The numbers are too small to permit multivariate assessment of the independent contribution of different variables (19).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Evaluation of persons exposed to newly reported cases of tuberculosis is undertaken with the dual goals of finding other new cases and also identifying persons with tuberculous infection who are candidates for preventive therapy. As a method of active case-finding, contact investigation is an effective strategy; overall, active tuberculosis in contacts accounted for 5% of cases of tuberculosis diagnosed in San Francisco during the study period, and 1% of the contacts investigated had tuberculosis. Surprisingly, DNA fingerprinting of M. tuberculosis demonstrated that 30% of contacts with tuberculosis developed the disease at nearly the same time as, but not as a result of transmission from, the index case.

The assumption underlying contact investigation is that M. tuberculosis is likely to have been transmitted from the index case, around whom the investigation is centered. However, frequently, contacts have other risk factors for tuberculous infection and disease, so the infection in the contact may well have come from a source other than the index case. This is particularly relevant for foreign-born contacts who come from high-prevalence countries. The presumed transmission link to the index case is especially important in inferring drug susceptibility patterns when no organism is available for susceptibility testing, as is always the case in latent tuberculous infection and, frequently, the situation in children with tuberculosis (20, 21). Although it is conceivable that an important proportion of contacts with tuberculous infection also have different strains that their index case, DNA-fingerprint comparison is only feasible for culture-positive cases, preventing confirmation of this possibility. Similarly, the majority (85%) of our tuberculosis cases in pediatric contacts were culture-negative, and thus our results cannot be generalized to this population.

In San Francisco, matched fingerprints were associated with the contact being born in the United States and being non-Asian. This observation is consistent with the finding that in general persons born in the United States have a much higher rate of clustering than do foreign-born persons (22). This is presumably because when exposed, a U.S.-born person is more likely to be vulnerable to the infection because of the absence of immunity provided by prior (latent) infection (23- 25) or possibly BCG vaccination. Conversely, foreign-born persons are more likely to develop active tuberculosis as a consequence of reactivation of prior infection.

Our findings demonstrate that in the management of contacts, the drug-susceptibility patterns of the organisms infecting the index case are generally, but not always, predictive of those infecting the contact. Not surprisingly, discordant susceptibility results were seen with greater frequency when the index case and contact had different strains of M. tuberculosis; in San Francisco this is more likely when the contact is foreign-born. Our findings confirm that inferred drug susceptibility can only be generalized to susceptible populations in whom recent transmission is likely and not to populations in which infection in contacts is less likely to be related to the index case. Hence, basing treatment recommendations on the drug-susceptibility pattern of the index case for U.S.-born contacts is a rational approach. This finding is especially relevant in treating severe forms of tuberculosis that are commonly culture-negative such as meningitis or pericarditis.

It is surprising that factors usually associated with enhanced transmission such as closeness of contact, cavitary disease in the index case, and HIV infection in the contact (not necessarily associated with transmission but rather progression from infection to active disease) were not associated with an increased frequency of matching fingerprints. This may be the result of insufficient statistical power to detect such associations since for each of these there was a slight trend towards an association with matched fingerprints.

The observation that 30% of cases identified through contact investigation did not acquire their infection from the putative source does not invalidate the utility of contact investigation as a public health intervention. In fact, 1.0% of all culture-positive cases diagnosed during the study period were identified through contact tracing but unrelated to the index case.