Latent TB Infection Treatment Acceptance and Completion in the United States and Canada FREE TO VIEW

Treatment of latent TB infection (LTBI) has been identified by the Centers for Disease Control and Prevention (CDC), the Advisory Council for the Elimination of Tuberculosis, the American Thoracic Society, and the Institute of Medicine of the National Academy of Sciences as the major strategy for elimination of TB in the United States.^1-3 It is estimated that 10 to 15 million people in the United States have LTBI, and roughly 300,000 Americans are treated for LTBI each year.^4,5 In both the United States and Canada, LTBI treatment is a cornerstone of national programs to reduce the burden of TB.

Factors that limit the effectiveness of this strategy are failure of persons with LTBI to accept or complete treatment. Although there are anecdotal reports that most persons offered treatment accept it, there have been no systematic studies of treatment acceptance. Individual clinics have reported completion by 5% to 60% of persons receiving the 6-month regimen of isoniazid (INH),^6-17 and 19% to 76% of persons receiving the 9-month INH regimen.^18-21 However, the data for the 9-month regimen were from subjects enrolled in research studies and may not reflect completion achieved under program conditions.

A more precise estimate of the proportion of persons declining and failing to complete treatment and identification of predictors of these outcomes would facilitate the development of interventions to improve acceptance and completion. In a previous study, we asked clinics to estimate the number of subjects treated for LTBI, but we were not able to determine acceptance or completion of such treatment. Therefore, in this study, we performed a randomized retrospective evaluation of the frequency of and risk factors for declining and failing to complete therapy for LTBI in the United States and Canada.

The study was conducted at the 19 sites of the Tuberculosis Epidemiologic Studies Consortium (TBESC), each of which represents a partnership between an academic institution and a state or local TB control program.⁴ Each TBESC site was asked to examine practice sites in their catchment area that provided LTBI treatment to at least 10 persons in 2002 and could provide access to records of patients.

The study design was a random stratified two-stage cluster sample survey with 1-year follow-up; all persons who were offered and/or started treatment of LTBI during 2002 at a participating clinic were eligible. This design permits adjustment for nonresponse and unequal selection probabilities in clinics of differing size. Outcomes were evaluated during 1 year of follow-up from the date of starting treatment. Patient charts were sampled using a random, stratified two-stage cluster sample survey. In the first stage, clinics were sampled across the sites, stratified by clinic type; in the second stage, charts were selected at random from within the clinics using predesignated sampling intervals. In addition to abstraction of individual patient charts, a clinic characteristic survey was performed by interviewing the clinic director or staff designated by the clinic director. The Institutional Review Board of the CDC determined that the study involved minimal risk to participants and approved waivers of Privacy Rule authorization and informed consent according to federal regulations.

Completion of LTBI treatment was based on a specified number of doses to be completed within a specified period of time for each regimen, as previously defined.¹ For the 9-month daily isoniazid regimen, completion was defined as 270 doses within 12 months; for 6 months of daily isoniazid completion was 180 doses within 9 months; for 4 months of daily rifampin, completion was 120 doses within 6 months; and for 2 months of daily rifampin/pyrazinamide, completion was 60 doses within 3 months.

The analysis considered two end points: acceptance and completion of LTBI treatment. We analyzed risk factors for failure to accept (ie, declining) treatment only in clinics where we were able to identify and sample subjects who had been offered treatment. The analysis of the second end point, treatment completion, was based on all patients started on treatment at any clinic. Frequencies and descriptive statistics used χ² tests and confidence intervals to assess statistical significance; logistic regression was used to identify and evaluate risks associated with declining and failing to complete treatment. Data analyses were weighted to adjust for nonresponse and unequal selection probabilities due to the two-stage sample design. The sampling weights were adjusted for nonparticipation of sampled clinics and for charts that were sampled within clinics but could not be located for abstraction. Weighted frequencies, proportions, other descriptive statistics, and logistic regression analyses were performed using WesVar (Westat; Rockville, MD).²² Multivariate models for declining and failing to complete treatment were constructed by including all variables associated with the outcome of interest that had a P value of .15 or less, and then removing variables by stepwise elimination to arrive at the most parsimonious model; a significant P value was defined as < .05.

Sixty-eight (68%) of the 100 clinics randomly selected for chart abstraction within the 19 catchment areas participated in the study. There were some differences between responding and nonresponding clinics. The response rate was substantially higher among public health clinics (82%) than among other types of clinics (26%); the latter group included private clinics, HIV clinics, and clinics serving homeless persons, immigrants, and refugees. There were also differences by size of clinic, with response rates of 50%, 75%, and 81%, respectively, among clinics reporting 100 or fewer, 100 to 249, and 250 or more LTBI patients. Finally, there were differences by region, with response rates of 76%, 59%, and 52% in the South, Northeast, and Midwest/West (combined). In summary, our sample somewhat overrepresents larger clinics, public health clinics, and clinics in the South. Because of the small sample size overall, we were unable to adjust for all these factors in weighting, but we did adjust for type of clinic, which had the greatest disparity in response rates. Reasons for failure to participate included inability to identify records of persons tested or treated for LTBI (5), administrative refusal to participate (8), failure to obtain local institutional review board approval (11), and other reasons (8). Of the clinics, 61/68 (90%) were general public health clinics, 2/68 (3%) were immigrant/refugee clinics, 2/68 (2%) were correctional clinics, one was an HIV clinic, one was a homeless clinic, and one was defined as other.

At the 68 clinics, 3,540 charts were selected for review, of which 2,395 (67.7%) were located and abstracted, 797 (22.5%) were ineligible, and 348 (9.8%) could not be located or were not abstracted for some other reason. Of the 68 clinics, only 32 both performed skin testing and offered treatment. Thus, only patients sampled at these 32 clinics who had testing done on site were included in the analysis of declining treatment.

Of 720 patients at the 32 clinics, 123 (17.1%; 95% CI, 5.8%-36%) declined initiation of treatment (Table 1). Of the 440 foreign-born persons, 232 (52.7%) were from Mexico, 46 (10.5%) were from the Philippines, 20 (4.5%) were from Vietnam, and 142 (32.3%) were from other countries. In multivariate analysis (Table 2), employees at a health-care facility were more likely to decline treatment, whereas contacts of a case of TB were less likely to decline. The proportion of persons declining treatment was not different between US/Canada-born and foreign-born persons, nor was it different between foreign-born persons in the United States/Canada for 5 years or less compared with those in the United States/Canada for more than 5 years. The proportion declining treatment was less among foreign-born persons in the United States/Canada 1 year or less (11.4%) compared with all others (25.7%), but this was not statistically significant (odds ratio, 0.373; 95% CI, 0.113-2.471; P = .464).

At the 68 clinics that provided treatment to those tested on site or elsewhere, 1,994 persons started treatment (Table 3). Of the 1,177 foreign-born persons, 482 (41.0%) were from Mexico, 146 (12.4%) were from the Philippines, 55 (4.7%) were from Vietnam, and 495 (42.0%) were from other countries, each of which represented less than 3% of the total. A regimen of 9 months of INH was prescribed for 1,674 (84.0%) of them, 6 months of INH for 181 (9.1%), 4 months of rifampin for 91 (4.6%), and 2 months of rifampin and pyrazinamide for 13 (0.7%). Overall, 1,054 (52.5%; 95% CI, 45.8%-59.2%) of the 1,994 patients did not complete treatment. Univariate analysis of characteristics associated with failure to complete treatment is shown in Table 3 and Figure 1. Of patients who were prescribed 9 months of INH but failed to complete the 9-month course, 21.9% failed to complete 6 months, whereas 78.1% completed 6 months but failed to complete the final 3 months. Thus, for the 9-month regimen, if failure had been defined as not taking at least 6 months of INH, the overall noncompletion rate would have been 42.9%.

In multivariate analysis, failure to complete therapy was associated with being a resident in a congregate setting, initiating the 9-month INH regimen compared with all others, being an injection drug user, age ≥ 15, and being an employee of a congregate setting (Table 2). As shown in Figure 2, more than 20% of persons initiating one of the three main LTBI treatment regimens stopped taking it in the first month. This was a significantly greater percentage than in all other months (P < .001).

Failure to complete treatment was not different between US/Canada-born and foreign-born persons, nor was it different between foreign-born persons in the United States/Canada for 5 years or less compared with those in the United States/Canada for greater than 5 years. However, failure to complete treatment was significantly less among foreign-born persons in the United States/Canada 1 year or less (48.0%), compared with all others (55.3%; relative risk, 0.824; 95% CI, 0.671-0.975; P = .0262).

This study shows that although acceptance of LTBI treatment is relatively high (83%), the proportion completing is low (47%). When the product of these two proportions is calculated, we see that completion was achieved by only 39% of persons who were offered LTBI treatment. The major risk factor for noncompletion was being prescribed the 9-month INH regimen. As shown in Figure 1, when the proportions completing the three most commonly prescribed regimens are compared, it can be seen that there is an inverse relationship between length of the regimen prescribed and the proportion completing treatment. The increased completion that we observed with the 4-month rifampin regimen is consistent with previous reports.^18-20 It is of interest to note that if completion of the 9-month regimen were defined as completion of at least 6 months of INH, the noncompletion rate of the 9-month regimen, 42.9%, would have been almost identical to that of the 6- month regimen, 44.8%. Thus, prescribing the longer course seems neither to greatly encourage nor discourage patients from completing 6 months of INH.

In addition to using regimens of shorter duration, another strategy for increasing completion suggested by the results of this study is targeting of populations at higher risk of failing to complete treatment. Such populations include residents in congregate settings (nursing homes, homeless shelters, jails, and prisons), injection drug users, and employees of hospitals and nursing homes. Some of these populations are recognized to include many difficult-to-reach subjects,¹ but health-care workers could be targeted by adherence programs in the workplace. Unfortunately, only 336/1994 (16.9%) of the patients receiving LTBI treatment in this study were members of these groups, so improvement in completion among these populations might not translate to a substantial overall improvement.

On the other hand, if strategies to improve completion in populations with low rates of completion cannot be implemented, focusing screening efforts on populations with better rates of completion might more efficiently use program resources. Such populations include those in contact with patients with infectious TB, persons younger than 15 years of age, and foreign-born persons during their first year in the United States/Canada. Such a strategy would have the additional advantage of targeting the group with the highest rates of TB disease of all foreign-born persons in the United States.^23-26 Moreover, specialized programs that are designed to be culturally sensitive have been shown to be successful in reaching such populations.²⁷

We did not find significant differences in completion of LTBI treatment by race, sex or birth outside the United States, or between those who had received bacillus Calmette-Guérin (BCG) vaccination and those who had not, as has been reported by others.^12,14,16,17 Moreover, characteristics of the treating clinic and the patient experience at the clinic were not associated with completion. Thus, factors that have been reported to be associated with completion in single-clinic studies may not be generalizable to the United States as a whole.

More than 20% of those initiating treatment stopped it within 1 month of starting; this amounts to one-third of all those who failed to complete treatment. It is possible that some of those failing to complete in the first month may have been individuals who were unwilling to refuse the offer of treatment, but who never intended to initiate it. Thus, acceptance rates may not be as encouraging as they appear. In addition, because more than one-half of those who did not complete stopped within 2 months after starting, close follow-up of patients during the first 2 months with prompt intervention to encourage completion among those stopping treatment might pay substantial dividends.

Our study has several limitations. First, not all clinics that were selected for the study agreed to participate, and clinics that did not participate differed from those that did participate. Although we were able to adjust for some of these differences in the analysis, our results may still reflect unmeasured or uncorrected bias. Second, clinics that treated fewer than 10 persons per year for LTBI were excluded. Thus, it is possible that large numbers of small clinics and practices, each treating a small number of persons each year, could represent a substantial number of LTBI diagnoses and courses of treatment that were missed by our survey. Based on our informal understanding of the relative importance assigned to LTBI treatment by primary care physicians and infectious diseases specialists, we doubt that this is the case. Third, some charts were ineligible or could not be located. Such charts might be expected to have been more likely to represent patients who declined or did not complete treatment. To the extent that this occurred, our proportions of acceptance and completion may be overestimates.

We conclude that the most important advance that could improve the effectiveness of LTBI treatment would be a regimen of shorter duration. The 4-month rifampin regimen and the 3-month regimen of daily INH plus rifampin are both shorter regimens with better completion rates, but the 4-month rifampin regimen has not had efficacy evaluated in a large prospective trial, whereas the 3-month regimen of INH plus rifampin has only been conclusively demonstrated to be effective in HIV-infected persons,²⁸ and is not recommended for use in the United States. Another 3-month regimen, once weekly rifapentine plus INH, has been shown to be well tolerated and lead to improved completion.²⁹ The efficacy of this regimen compared with the 9-month INH regimen is currently being studied by the Tuberculosis Trials Consortium.³⁰ If this regimen is as effective as the 9-month INH regimen, its widespread adoption might lead to substantially improved treatment completion rates and result in prevention of more cases of TB.

Author contributions:Dr Horsburgh Jr: contributed to the design of the study, abstraction of study data, and analysis of the data, and wrote the first draft of the manuscript.

Dr Goldberg: contributed to the design of the study, analysis of the data, and preparation of the manuscript.

Dr Bethel: contributed to the design of the study, analysis of the data, and preparation of the manuscript. He had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Dr Chen: contributed to the creation of the study database, analysis of the data, and preparation of the manuscript.

Dr Colson: contributed to the design of the study, abstraction of study data, analysis of the data, and preparation of the manuscript.

Ms Hirsch-Moverman: contributed to the design of the study, abstraction of study data, analysis of the data, and preparation of the manuscript.

Dr Hughes: contributed to the design of the study, abstraction of study data, analysis of the data, and preparation of the manuscript.

Ms Shrestha-Kuwahara: contributed to the design of the study, analysis of the data, and preparation of the manuscript.

Dr Sterling: contributed to the design of the study, abstraction of study data, analysis of the data, and preparation of the manuscript.

Ms Wall: contributed to the design of the study, abstraction of study data, analysis of the data, and preparation of the manuscript.

Mr Weinfurter: contributed to the design of the study, analysis of the data, and preparation of the manuscript.

Financial/nonfinancial disclosures: The authors have reported to CHEST the following conflicts of interest: Dr Sterling has received grant money from Pfizer and Oxford/Immunotec. Mr Weinfurter has a sibling who works for and owns stock in Amgen. Drs Horsburgh, Goldberg, Bethel, Chen, Colson, and Hughes, and Mss Hirsch-Moverman, Shrestha-Kuwahara, and Wall have reported that no potential conflicts exist with any companies/organizations whose products or services may be discussed in this article.

Role of sponsor: The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Other contributions: We thank the investigators and staff at all the participating TBESC sites: James McAuley, MD, MPH, Judith Beison, Chicago, IL; Frank Wilson, MD, Cheryl LeDoux, MT, MPH, Little Rock, AR; Jennifer Flood, MD, MPH, Sumi Sun, MPH, Berkeley, CA; Hugo Ortega, New York, NY; Randall Reves, MD, MSc, Denver, CO; Henry M. Blumberg, MD, Jane Tapia, BSN, Atlanta, GA; Jessie Wing, MD, Sara Jacobson, MPH, Honolulu HI; Cara Endyke-Doran, RN, BSN, MPH, Baltimore MD; Sue Etkind, RN, MPH, Sharon Sharnprapai, MPH, Boston MA; Wendy Mills Sutherland, MPH, Hodan Guled, MPH, Minneapolis, MN; John Grabau, PhD, MPH, Wilson Miranda, MBBS, MPH, Albany, NY; Rachel Royce, PhD, Carol Dukes-Hamilton, MD, Juani Munoz Sanchez, Durham, NC; Connie Haley, MD, MPH, Tamara Chavez-Lindell, MPH, Nashville, TN; Edward Graviss, PhD, MPH, Smita Chatterjee, MS, Houston, TX; David E. Griffith, MD, Tyler, TX; Michael Kimerling, MD, MPH, Ashutosh Tamhane, MD, DPH, MSPH, Birmingham, AL; Monika Naus, MD, MSc, Mark Fitzgerald, MD, DM, Victoria Cook, MD, Maya Nakajima, BSc, Vancouver, BC; Earl Hershfield, MD, Barbara Roche, BScN, MPH, Winnipeg, MB; Nandini Selvam, MPH, PhD, Newark, NJ; Stephen Weis, MD, Guadalupe Munguia, MD, MPH, Jingsheng Yan, PhD, MPH, Heidi L. Venegas, MS, Sarah Brown, MS, Fort Worth, TX; Mike Jones, MS, Bethesda MD. In addition, we thank Richard O’Brien, MD, for inspiration and encouragement in the development of this study.