Multidrug Resistance Among Persons With Tuberculosis in California, 1994-2003 FREE

California has made significant advances in controlling tuberculosis (TB), as exhibited by a 33% decrease in notified cases (4834 to 3224) during 1994 to 2003.¹ Tuberculosis nevertheless continues to have a substantial public health impact, leading in 2003 to 233 deaths and more than 132 cases diagnosed among children younger than 5 years. Additionally, cases of TB due to strains of Mycobacterium tuberculosis that are resistant to the mainstay first-line drugs isoniazid and rifampin (ie, multidrug-resistant [MDR] strains) continue to appear in California despite high rates of treatment success.

The emergence of such cases, requiring prolonged treatment for at least 18 months and exhibiting higher rates of treatment failure and poorer outcomes, threatens the efficacy of TB control efforts.^2- 10

Case fatality in multidrug resistant tuberculosis (MDR-TB) ranges from 12% among individuals without human immunodeficiency virus (HIV) infection to as high as 90% among persons infected with HIV.¹¹ Moreover, managing patients with drug resistance requires considerable expertise and resources; health care cost estimates for individual MDR-TB patients in the United States range from $28 217 to $1 278 066.¹² Finally, MDR-TB has also been associated with serious sizeable nosocomial and community outbreaks in California and the greater United States.^5,13- 18

Through cooperation with local health departments, the California Department of Health Services Tuberculosis Control Branch has collected data since 1985 on the numbers of reported TB cases for the purpose of state and national surveillance. Nationally, during the late 1980s and early 1990s, a resurgence of TB disease and M tuberculosis strains resistant to multiple anti-TB drugs seriously threatened TB control efforts. In response, the US Centers for Disease Control and Prevention conducted drug resistance surveys in 1991 and 1992 and in 1993 added susceptibility results to the information collected by the national surveillance system.^4,19 Since 1994, California has collected population-based drug susceptibility results for M tuberculosis isolates. To better understand the impact of resistance to multiple drugs on TB control in California and to plan public health interventions, we analyzed drug susceptibility data in the California TB surveillance system to describe the magnitude, trends, geographic distribution, clinical characteristics, risk factors, and outcomes of drug-resistant TB cases.

We analyzed data for TB cases submitted during 1994-2003 to the California Department of Health Services’ registry of Reports of Verified Cases of TB (RVCT). We included M tuberculosis culture-positive cases with initial drug susceptibility test results for at least isoniazid and rifampin. To describe the association between MDR-TB and prior TB treatment in general, we used the RVCT variable describing previous TB history. To describe the association between MDR-TB and prior treatment in California, we queried the TB registry of preceding years (1985-1993) for evidence that a case reported in the study period had previously received TB treatment in the state.

Since the RVCT does not include HIV status, the case registry was matched with the State of California AIDS case registry for 1994-2002. Immigrants identified as having smear-positive pulmonary TB during overseas screening examinations are classified in the A category and may be permitted to immigrate if TB therapy results in sputum conversion; those with radiological abnormalities consistent with TB who are smear-negative are given class B status. We crossmatched MDR-TB cases with the California immigration A/B notification registry, which was only available for 1998-2003. Analyses considering completion of therapy were conducted only on cases reported between 1995 and 2001. Incidence rates for foreign-born population groups were computed using California Department of Finance population estimates.^20- 21 The MDR-TB midyear point prevalence was determined using start and stop dates of TB drug therapy.

Categorical data were compared by the χ² test. The Wilcoxon rank sum test was performed to determine differences in distributions of continuous variables. The Cochran-Armitage test was used to identify trends. P values less than .05 were considered statistically significant. Univariate analyses were performed with SAS version 8.0 (SAS Institute Inc, Cary, NC).

We conducted multivariable analyses to compute odds ratios (ORs) adjusted for other covariates. Commonly used methods (eg, stepwise regression) do not consider the effect of variable selection and for this reason may yield biased estimates of effect size as well as misleadingly small P values.^22- 23 We used the method of Bayesian model averaging,²² since this method yields adjusted ORs that take into account the effect of variable selection; the multivariable analyses themselves were conducted in S-Plus version 3.4 (Insightful Corp, Seattle, Wash) using the program bic.glm.²⁴ This method yields adjusted ORs, approximate confidence intervals (CIs), and posterior probabilities that the OR corresponding to each correlate differs from the null value of 1; values greater than 50% support the statement that the particular correlate corresponds to a risk factor.²² Specifically, we used the following variables as correlates for MDR status: previous TB diagnosis, positive acid-fast bacilli sputum smear results, abnormal radiography findings, age, sex, race/ethnicity (self-reported as part of routine surveillance data, using categories determined by the US Census Bureau), health care provider type, type of TB drug therapy (self-administered therapy, directly observed therapy, or both), homelessness, treatment outcome, and length of time in the United States for foreign-born cases.

Of 38 291 TB cases reported in California between January 1, 1994, and December 31, 2003, 29 393 (77%) had a culture positive for M tuberculosis, and 28 712 (75%) were tested for resistance to at least isoniazid and rifampin. Of these, 407 (1.4%) were MDR. The sociodemographic characteristics of the 28 712 cases are shown in Table 1. The number of incident MDR-TB cases decreased during the 9-year period, from 57 in 1994 to 33 in 2003 (Figure 1). Although there was also a 33% decrease in the number of annually reported TB cases over the study period, the proportion of cases with MDR-TB did not significantly change (P = .87). Additionally, the proportion of isoniazid resistance among those with no reported history of TB did not significantly change over the study period (P = .33) (Figure 2). The mean annual incidence of MDR-TB was 0.13 and ranged from 0.09 to 0.18 per 100 000 population. The median MDR-TB point prevalence was 60 cases (range, 50-73).

Cases of MDR-TB were found in 38 (62%) of California’s 61 health jurisdictions and were reported both from large urban areas and from smaller health jurisdictions with less experience managing TB. Health jurisdictions reporting an average of 30 or fewer TB cases per year over the study period reported 9% of MDR-TB cases. In these jurisdictions with lower TB morbidity, reporting of MDR-TB increased from 7% of all MDR-TB cases in 1994 to 21% in 2003 (P = .39).

Of the 407 MDR-TB cases, 367 (90%) were pulmonary (Table 2). Smear-positive TB cases were nearly twice as likely to be MDR than were smear-negative cases (adjusted OR, 1.71; 95% CI, 1.30-2.26) (Table 3). Cases of MDR-TB were twice as likely to have cavitary lesions compared with non–MDR-TB cases (P<.001), and having a cavitary lesion was significantly associated with being smear-positive (P<.001). Additionally, MDR-TB cases were 7 times more likely to have reported previous treatment for TB (P<.001). Of all MDR-TB cases, 124 (31%) had had a prior episode of TB according to the RVCT. Of these, 27 (22%) had been treated for previous TB in California, and nearly half of these demonstrated increased resistance from the previous TB episode to the study episode. Tuberculosis cases with HIV/AIDS were significantly less likely to have MDR-TB compared with those without HIV/AIDS (P = .01).

Of the 407 MDR-TB cases, 280 (69%) occurred in persons who reported no prior history of TB, and the proportion of MDR-TB cases without a history of prior TB treatment increased from 65% (n = 37) in 1994 to 71% (n = 22) in 2003, though this increase was not statistically significant (P = .89). The adjusted OR for reported previous diagnosis as a risk factor was 7.02 (95% CI, 5.37-9.18) (Table 3). Patients resistant to either isoniazid or rifampin may develop resistance to the other drugs and thus become classified as MDR cases. Of all culture-positive cases reported from 1994-2003, 2665 (9%) had isolates that were not MDR but were resistant to at least isoniazid or rifampin, and this proportion did not change during the study period (P = .65). Resistance to at least isoniazid among persons without a known history of TB ranged from 8% to 10% and did not increase during the study period (P = .33). Resistance to rifampin only averaged 0.4% per year for the study period, and TB cases with AIDS were nearly 7 times more likely to have rifampin-only resistance compared with those without AIDS (35 [1.7%] vs 67 [0.3%]; relative risk, 6.97; 95% CI, 4.62-10.51).

Cases of MDR-TB were younger than other TB cases (mean age, 43 vs 48 years; P<.001 (Table 1); adjusted OR for the age category 65 years and older, 0.36; 95% CI, 0.25-0.53) (Table 3). Of all MDR-TB cases, 6 (2%) were children younger than 5 years. Non–US-born cases were twice as likely to be MDR compared with US-born cases (P<.001). Of non–US-born MDR-TB cases, 113 (35%) were among persons who had been in the United States for less than 1 year at diagnosis (median, 3.5 years). At the time of case report, non–US-born MDR-TB cases had been in the United States for a shorter mean duration than other TB cases (6.7 vs 12.1 years, P<.001); the adjusted OR for MDR-TB (with US-born as baseline) was 2.47 (95% CI, 1.34-4.55) for those in the United States for less than 1 year at the time of diagnosis. Mexico, the Philippines, Laos, Vietnam, and South Korea were the most common countries of origin of MDR-TB cases originating outside the United States (Table 4). The incidence of MDR-TB among foreign-born cases (3.5 per 100 000 population) was significantly higher than the 0.2 per 100 000 among US-born cases (P<.001). The 5-year A/B registry crossmatch found that only 19 (10%) of foreign-born MDR-TB cases arrived in the United States with an A/B notification following overseas screening. Of these, 4 of 13 class A classifications had MDR-TB and the remaining 15 MDR-TB cases arrived with a class B classification.

Cases of MDR-TB were less likely to report excessive alcohol use or homelessness compared with non–MDR-TB cases (P<.001 and P = .02, respectively) (Table 1). Similarly, MDR-TB was less associated with injection drug use, residence in a long-term care facility, incarceration, or high-risk occupation (eg, health care, correction, migratory agricultural worker). However, the adjusted OR for homelessness did not differ from 1.

Of the 346 MDR-TB cases with treatment outcome data (85%), 231 (67%) completed therapy, 49 (14%) died, 48 (14%) moved, 9 (3%) were lost to follow-up, 1 refused therapy, and 8 (2%) had an unspecified outcome (Table 2). Cases with MDR-TB isolates were significantly less likely to complete therapy when compared with those without MDR-TB (P<.001). Cases with MDR-TB who died during TB treatment were significantly younger (mean age, 53 vs 63 years; P = .001) but were not significantly more likely to have AIDS (P = .56).

Significantly fewer MDR-TB cases were managed exclusively by private physicians compared with non–MDR-TB cases (50 [14%] vs 9674 [35%], P<.001) (Table 2). The proportion of MDR cases managed exclusively by county health departments increased from 48% in 1994 to 59% in 2001 (P = .02). Although MDR-TB cases were more likely to receive directly observed therapy (DOT) compared with non-MDR cases (P<.001), 52 (15%) of 345 MDR-TB cases with available therapy data were reported as receiving only self-administered therapy. Patients cared for by private physicians were less likely to receive DOT when compared with those managed by public providers (34% vs 60%; relative risk, 0.35; 95% CI, 0.18-0.70; P = .002). However, patients with MDR-TB managed by private physicians were just as likely to complete therapy (56% vs 68%; relative risk, 0.60; 95% CI, 0.31-1.17; P = .16).

Of the 407 cases with MDR, 71 (17%) were resistant to only isoniazid and rifampin and 86 (21%) were resistant to at least all 4 first-line drugs (Table 5). Although reporting of test results for fluoroquinolone resistance was limited to 236 (58%) MDR-TB cases, 28 (12%) were resistant to fluoroquinolone(s) and this proportion did not increase over the study period. Resistance to injectable TB drugs occurred in 62% of all MDR-TB cases and this proportion did not significantly change over the study period.

We used the following variables in the multivariable model: reported history of previous TB, smear status, chest radiography results, age category, sex, race/ethnicity, provider type, type of therapy (eg, directly observed or self-administered), length of time in the United States for non–US-born cases (categorized as indicated), homelessness, treatment outcome, and AIDS. Previous TB diagnosis, positive acid-fast bacilli sputum smear results, Asian/Pacific Islander ethnicity, time in the United States of less than 5 years at the time of TB report, as well as the outcomes of “died” and “moved” were identified by multivariable analysis as independent correlates positively associated with MDR-TB (Table 3).

In California, MDR-TB, a life-threatening airborne disease with expensive and limited treatment options, continues to persist despite current TB control efforts. Although the annual number of TB cases decreased 33% over the study period, the proportion of MDR-TB remained at approximately 1.4%. Additionally, we found that 83% of MDR-TB cases had resistance to other drugs in addition to isoniazid and rifampin. Cases of MDR-TB not only occurred in large urban jurisdictions, but we found an increasing proportion of cases in rural or smaller jurisdictions where staffing, resources, and TB expertise may be less available. Our findings are of concern and suggest that the cases of MDR-TB in California may have appeared for any of 3 reasons, which will be discussed in turn: importation of MDR strains from outside the state, endogenous development of MDR strains due to inadequate case management or poor treatment within California, or ongoing transmission.

We found that MDR-TB was strongly associated with birth outside the United States and that 83% of MDR-TB cases were foreign born, from 30 different countries. Population-based surveys in the 3 most common countries of origin (Mexico, the Philippines, and Vietnam) for individuals born outside the United States with TB in California have demonstrated levels of drug resistance that are 2 to 4 times that in the United States.^25- 27 We also found disproportionately high representation in individuals from Laos, the former Soviet Union, Korea, and Peru. While these findings highlight the value to California of international TB control measures such as DOT, only 19 persons (10%) with MDR-TB were identified through preimmigration screening, despite the fact that nearly one third of non–US-born MDR-TB cases were diagnosed within 1 year of arrival. Most people (such as students, workers, or undocumented immigrants) entering the United States from countries with high TB rates enter without TB screening at all, and even the mandated screening program for legal immigrants does not include sputum culture or susceptibility testing. As a result, MDR-TB is often not identified until diagnosed in the United States.

While TB rates among US-born persons have decreased, rates of TB and MDR-TB among minorities remain significantly elevated.^25- 27 Additionally, many risk factors previously observed to be associated with TB outbreaks among US-born persons (eg, homelessness, drug use, and AIDS) were found less frequently among MDR-TB cases in California compared with non–MDR-TB cases.^13- 17 Several MDR-TB outbreaks in the early 1990s that occurred in the eastern United States were associated with HIV, drug use, and health care workers.^28- 29 However, we found no association with drug use or health care workers, and AIDS was twice as likely to be associated with non-MDR-TB compared with MDR-TB. These findings suggest that considerable geographic differences may exist in TB and MDR-TB epidemiology in the United States and that these differences may change over time.

We found that MDR-TB was strongly associated with a reported history of previous TB treatment. A reported history of previous TB treatment, although unconfirmed, suggests that MDR-TB was acquired during a previous treatment episode; such acquired drug resistance may indicate failure of TB control efforts due to inadequate case management, interruptions in drug supply, or inadequate drug regimens. Although MDR-TB may have been acquired in the United States, the majority of non–US-born MDR-TB cases in our study reported a history of prior TB treatment prior to arrival in the United States. In addition to the possibility of importing MDR strains from abroad, our crossmatch with the historical California TB registry suggests that, while comparatively uncommon, some MDR cases arise as the result of inadequate case management or treatment in California. This risk is highlighted by our finding that nearly 10% of TB cases were reported as being resistant to either isoniazid or rifampin, thus placing them at higher risk of developing MDR-TB.

Finally, MDR cases may appear because individuals with MDR disease transmit the infection to others. Recent molecular epidemiology data suggest that transmission of MDR-TB is occurring in multiple sites in the United States, including California, and may be responsible for as many as one third of MDR-TB cases (unpublished data). Findings from our study revealed that MDR-TB cases are much more likely to have smear-positive cavitary pulmonary disease, which may contribute to the spread of drug-resistant strains. Additionally, the mean time to culture conversion for MDR-TB cases was 46 days longer than for non–MDR-TB cases (P<.001; data not shown). Patients with MDR-TB often do not start effective treatment until susceptibility results are known, which may also prolong the infectious period. The median point prevalence in California during the study was 60 cases. The prolonged 18 to 24 months of therapy translates into high management burdens for patients, physicians, and health jurisdictions and may also provide further opportunities for transmission in cases that do not respond to treatment. Finally, a higher proportion of MDR-TB cases moved or were lost to follow-up, which may also contribute to ongoing MDR-TB transmission.

Turning from the origin of MDR cases to their outcomes, our study has also shown that MDR-TB cases were significantly less likely to complete treatment and were more likely to die than those with non–MDR-TB.¹⁰ This association between resistance and poor outcome held true regardless of history of prior TB; a diagnosis of AIDS was associated with even poorer outcomes, including a higher likelihood of death. An increasing proportion of the MDR-TB cases were managed by the health department, and although private physicians yielded similar outcomes, they were significantly less likely to use DOT.

Several limitations apply to our findings. Only 75% of reported TB cases were culture confirmed and had drug susceptibility test results for isoniazid and rifampin. However, the absence of significant differences between persons with culture-positive TB with isolates available for testing and those without isolates suggests that our sample is representative of TB cases reported during the study period. Also, incidence rates by country of birth should be interpreted with caution because population estimates may not accurately capture data for migratory populations. The registry of immigrants with A/B status has only 5 years of data and the majority of non–US-born persons are not screened with this system when entering the United States. Treatment history is very difficult to determine and, although our results associating history of treatment with MDR-TB are plausible, they should be treated with caution. Finally, it is possible that much of the MDR-TB reported among non–US-born persons was acquired in their country of origin; however, in the absence of more detailed information we cannot determine the contribution of ongoing community transmission within the United States.

The findings of our study have several clear implications for TB control efforts. First, the fact that the majority of MDR-TB cases were foreign born highlights both the importance of international TB control (prevention of MDR development and transmission abroad) as well as the need to expand overseas screening programs to encompass additional high-risk groups, coupled with measures to ensure timely detection and treatment of MDR-TB once it develops.³⁰ Second, our results suggest that adherence to recommended TB treatment guidelines must be improved to ensure that poor case management does not contribute to further cases of MDR within California. Third, the higher proportion of individuals moving or lost to follow-up, as well as the longer time to culture conversion and clinical characteristics favoring transmission, suggest that measures to reduce transmission and improve outcomes are also necessary. Fourth, additional resources (eg, additional staff, regional centers of excellence, and “warm lines” that provide clinical consultations) are needed because an increasingly large proportion of MDR cases appear to be arising in rural or smaller health jurisdictions with limited resources and expertise; the threat of MDR-TB is exacerbated by a shrinking pool of clinicians experienced in managing these complex patients, who require intensive monitoring (eg, drug levels, second-line drug susceptibilities, and renal function) over an 18- to 24-month period. To help support the efforts of local programs to manage patients with complex MDR-TB, the California Department of Health Services TB Control Branch established an MDR-TB clinical service that provides clinical support, collaborates with model centers, and will participate in the efforts of the Centers for Disease Control and Prevention to support several TB consultation medical training centers. Our study suggests that clinicians should consider MDR-TB in younger persons with TB who are Asian and/or Pacific Islander, non–US-born from countries with known MDR-TB epidemics,³⁰ recent arrivals (<5 years) in the United States, and those reporting prior TB treatment.

Multidrug-resistant TB requires complex management decisions, and additional resources will be required to successfully interrupt transmission and cure patients through timely diagnosis, treatment with adequate drug regimens and DOT, and through a patient-centered approach to ensure adherence. Although MDR-TB may be curable at a great individual and societal cost, the implementation of both local and global TB control strategies is needed to prevent the further development and spread of MDR-TB.