Approval for this study was obtained from the University of Texas Health Center at Tyler (UTHCT) Human Subjects Institutional Review Board. Susceptibility results of the Mycobacteria/Nocardia Research Laboratory at UTHCT were screened for clarithromycin resistance among patients with MAC treated at UTHCT between 1991 and 2005. All patients with resistant isolates had multiple positive sputum specimens that grew MAC, abnormal chest radiographs, and consistent clinical findings as defined by the most recent American Thoracic Society statement on nontuberculous mycobacteria (8). None was known to be seropositive for HIV or at high risk for HIV infection, and all had MAC isolates with clarithromycin minimum inhibitory concentration (MIC) of 32 μg/ml or greater (9, 10). Patients with cystic fibrosis were excluded.

Approximately 50% of patients with macrolide-resistant MAC were enrolled in one or more noncomparative, open, prospective treatment trials that involved either clarithromycin or azithromycin (1–6). Patients in the trials signed an informed consent approved by the UTHCT institutional review board. Patients received their medications either daily or three times weekly. After completion of these trials, all subsequently encountered patients with initial macrolide-susceptible MAC lung disease were treated either daily or with the three-times-weekly regimen with a three-drug regimen using the drug dosages used in the clinical trials but without being part of a treatment study. The remaining 24 patients with macrolide-resistant MAC were referred to one of the authors because of treatment failure and/or their MAC isolate had become macrolide resistant.

Patient histories were reviewed for risk factors that may have contributed to the development of macrolide resistance. These included macrolide monotherapy, macrolide plus only a quinolone, and a macrolide combination regimen other than a quinolone alone that did not include ethambutol (e.g., a macrolide plus rifampin or rifabutin). These were considered as potential risk factors if given for 1 mo or more. Noncompliance was considered to be present if the patient did not take his/her medicines as instructed (11).

The details of the medical and surgical therapy of patients after their isolates were found to be macrolide resistant were reviewed. All patients were treated at UTHCT by one of the authors. No written treatment protocol was followed, but each case was discussed among the authors. Surgical resection for cure or debulking of the patients' MAC lung disease was considered in each patient. The chest surgeons involved in these cases had more than 30 yr experience with mycobacterial surgery, including MAC lung disease (12). Surgical results for four of these patients have been published previously (12).

Drug therapy used for patients with macrolide-resistant MAC, especially over the past 8 to 10 yr, included discontinuation of the macrolide, and initiation of therapy with ethambutol 25 mg/kg daily, rifabutin 300 to 450 mg daily, and streptomycin 5 to 10 mg/kg (up to 1,000 mg daily), or amikacin 5 to 7 mg/kg (up to 500 mg once daily). In patients with normal serum creatinine and blood urea nitrogen and an age younger than 60 yr, the injectables were usually given five times per week for the first 3 mo, then decreased to three times weekly for as long as could be tolerated. Patients between the ages of 60 and 70 were dosed three times weekly, and those older than 70 were dosed only twice weekly, again for as long as could be tolerated. With the exception of two patients, injectables were given intramuscularly. Aminoglycoside serum levels were not routinely obtained, in part because of the difficulty in obtaining streptomycin serum levels. Patients were treated until they were culture negative for 12 mo on therapy.

Aminoglycoside-related ototoxicity was monitored clinically by questioning for tinnitus and subjective changes in hearing at baseline and on each clinic visit. Audiograms were done if such changes were reported. Aminoglycosides were continued for the duration of therapy, if possible, in patients who had conversion of sputum to acid-fast bacilli (AFB) culture-negative status. Drug dosages or frequency were not changed for the development of tinnitus, but were changed or the aminoglycoside discontinued on an individual basis after review with the patient when subjective hearing loss was confirmed by audiogram. Other safety tests, including monitoring of renal and hepatic function, and blood counts were performed routinely. Patients filled out a symptom questionnaire at the time of each visit. Ethambutol-related toxicity was monitored with eye examinations, including the Snellen eye chart and the Ishihara color book, performed at baseline at the time of each visit, and whenever new eye symptoms (blurriness, change in vision) occurred (13).

Sputum samples for AFB analysis were obtained on a monthly basis when possible while on therapy and at least every 2 mo off therapy for the first 12 mo. Details of sputum processing and culture methods have previously been published (1, 3, 14).

Sputum conversion was defined as a minimum of three consecutive negative cultures over a minimum time of 3 mo. Cure was defined as negative sputum cultures for the infecting MAC strain for 12 mo or longer off drug therapy. Relapse isolates were compared with initial strains using pulsed-field gel electrophoresis (PFGE) (14–16).

Isolates were submitted to the Mycobacteria/Nocardia Research Laboratory at UTHCT. Organisms from a swab of an isolation plate were placed in tryptic soy broth with 15% glycerol and frozen at −70°C until needed. Organisms were identified as MAC based on a commercial DNA/RNA probe (AccuProbe; GenProbe, Inc., San Diego, CA). Macrolide-resistant MAC isolates were also identified to species using restriction fragment length polymorphism (RFLP) analysis of the hsp65 gene (see PCR Restriction Enzyme Analysis of the hsp65 Gene) (17, 18).

Susceptibility testing to clarithromycin was performed by broth microdilution in Mueller Hinton broth supplemented with oleic acid, albumin, and dextrose using current Clinical Laboratory Standards Institute guidelines (10). Isolates with MICs of 32 μg/ml or greater were considered resistant (10). MICs for azithromycin were not performed, with clarithromycin used as the class drug for the macrolides (19, 20).

PFGE was performed on macrolide-susceptible and -resistant isolates from the same patient (where available) using previously described techniques (14–16). Definitions of isolates as the same (indistinguishable), probably related, possibly related, or unrelated (different or unique) were the same as in a previous MAC study (14). These were modified from the criteria of Tenover and colleagues used to define bacterial outbreaks (21).

Genomic DNA was isolated; polymerase chain reaction (PCR) was then performed on a DNA fragment of approximately 445 bp in the 23S r-RNA that spanned position 2058 and 2059. PCR product purification was performed by the enzymatic ExonucleaseI/Shrimp Alkaline Phosphatase method. All sequencing reactions had excess dyes removed by size exclusion spin columns (Millipore, Billerica, MA) and were electrophoresed on an ABI 377 DNA sequencer according to Applied Biosystems (Big Dye Terminator; Applied Biosystems, Foster City, CA) standard sequencing protocols. Sequencing was performed by Amplicon Express (Pullman, WA).

A 441-bp portion of the hsp65 gene (Telenti fragment) was amplified by PCR as previously described (17, 18), then digested with BstEII or HaeIII. PCR restriction enzyme analysis (PRA) patterns were compared with the previous profiles characterized by Smole and colleagues (18) for M. avium, M. intracellulare, and undefined MAC taxa.

Group data are expressed as means ± SD. Comparison of outcomes between patient treatment groups was done with Fisher's exact test. Analysis of the effect of clinical variables on treatment response was done with the t test for equality of means after evaluation of the data with Levene's test for equality of variances. Two-tailed p values were used for all t tests. Significance of all comparisons was determined with a p value of less than 0.05.

Characteristics of all 51 patients with macrolide-resistant MAC isolates are shown in Table 1. Of the 27 patients with upper lobe cavitary disease, the majority were males (78%), most were former heavy smokers, one-third abused alcohol, and their mean age at the onset of MAC disease was 58 yr. These are typical features of patients with MAC with upper lobe cavitary disease (22–24). Of the 24 patients with nodular disease, more than 90% were women, the majority (71%) had never smoked, and none abused alcohol. Their mean age at the onset of MAC disease was 65 yr. These features are typical of patients with nodular type MAC lung disease (25–28).

The 51 patients were identified as part of drug treatment trials, routine drug therapy, or a referral because of failed drug therapy. A total of 362 patients were enrolled in one or more of seven multidrug macrolide treatment trials between 1991 and 2000 (1–6). Twenty-four patients (6.6% of the total) enrolled into trials initially had macrolide-susceptible MAC isolates and subsequently MAC isolates became clarithromycin resistant.

Resistance among patients in the drug trials was identified in two treatment groups (see Table 2). Group 1 included 59 patients who received 4 mo of initial macrolide monotherapy, with subsequent addition of ethambutol and a rifamycin (1). All isolates of MAC were susceptible when treatment was initiated, but 12 isolates (20.3%) became resistant over the subsequent months of therapy.

Group 2 consisted of 303 patients who received three-drug therapy from treatment initiation, but with different macrolides, rifamycins, or dosing intervals. Only two isolates (0.7%) developed resistance to macrolides within the first 6 mo of therapy, and only 12 isolates of the 303 (4.0%) became resistant throughout the treatment period regardless of whether the patients were drug responders (i.e., sputum converted to negative) or nonresponders, compliant with therapy, or tolerant of the macrolide companion drugs. The incidence of macrolide resistance was 6 of 170 (3.5%) in patients receiving clarithromycin, 5 of 133 (3.8%) in patients receiving azithromycin, 6 of 141 (4.25%) in patients receiving daily therapy, and 5 of 162 (3.0%) in patients receiving intermittent therapy.

An additional 24 patients were referred for treatment due to previous treatment failure or because of known macrolide resistance, and their isolates were confirmed to be macrolide resistant.

An additional three patients placed on an initial three-drug regimen but not part of treatment trials (two patients) or an IFN-γ randomized trial (one patient) also had MAC isolates that became macrolide resistant.

Initial macrolide monotherapy, a macrolide plus a quinolone, or either of these occurring as a result of discontinuation of ethambutol appeared to be the major reasons for development of resistance in most patients. These reasons changed with time. In the first 5 yr (1991–1995), all resistance (17/17 cases) was related to macrolide monotherapy. Over the subsequent 10 yr (1996–2005), this became less common (only 11/34, or 33%), with the use of a macrolide plus a quinolone (essentially monotherapy) becoming increasingly important (11/34, or 33%). Overall, 39 of 51 (76%) isolates from patients with known histories became resistant after one of these two therapeutic approaches. Only 9 of 51 isolates (18%) from patients with available histories from 1991 to 2005 developed macrolide resistance when started on standard three-drug therapy.

The 51 patients were divided into four groups on the basis of their therapy for macrolide-resistant MAC (see Table 3). Of the 14 patients from group 1 who had lung resections and prolonged aminoglycoside therapy (mean, 12.0 ± 4.1 mo; range, 7–19 mo) combined with daily high-dose ethambutol and rifabutin, sputum samples from 11 patients (79%) have converted to culture negative. Nine of these patients have completed therapy, one is still on therapy, and one died of respiratory complications. These latter two patients have had more than 6 mo of negative sputum cultures. Of the patients included as treatment failures, two died of medical complications within 6 wk of their lung surgery, so the microbiological success of the surgery could not be assessed. The third patient considered a failure was lost to follow-up after surgery in another institution. Thus, of the patients who survived their perioperative period with adequate follow-up, 11 of 11 (100%) are culture negative to date. No patient who met the case definition of sputum conversion in this group has relapsed on or off therapy.

Of the nine patients in group 1 whose sputum sample converted and who are off drug therapy and considered cured, two have died of diseases unrelated to MAC or their surgery and the remainder are still alive with a mean follow-up of 28.6 ± 13.7 mo (range, 12–48 mo) since sputum conversion. Overall, of the 14 patients from group 1, 5 (36%) have died, with a mean survival of the nonsurgical mortality of 31.0 ± 31.4 mo (range, 3–62 mo).

Of the remaining 37 patients, 2 received surgery alone without prolonged (⩾ 6 mo) aminoglycoside therapy (group 2), 8 received prolonged aminoglycoside therapy without surgical resection (group 3), and 27 patients received neither of these treatment modalities (group 4). Only two patients (5.4%) from these three groups converted their sputum specimens to negative (see Table 3). Of these 37 patients, 22 (59%) are known to have died of this disease, with a mean survival of 23.8 ± 25.5 mo (range, 2–81 mo). Of the two patients from groups 3 and 4 who converted their sputum to negative, one has relapsed and the other has been lost to follow-up.

Of the patients who failed therapy (totals, groups 2–4), the 1-yr mortality from diagnosis of their MAC resistance was 13 of 38 (34%) and the 2-yr mortality was 17 of 38 (45%). Eight of these patients survived 5 yr after the diagnosis of macrolide resistance; 11 are still living but have survived less than 5 yr (follow-up, 16–58 mo).

Of the patients whose sputum converted to negative, the 1- and 2-yr mortality was 0 of 13 (0%), with the caveat that some patients in both groups have been monitored less than 2 yr. Two of these patients survived 5 yr after the diagnosis of macrolide resistance; 10 are still living but have survived less than 5 yr (follow-up, 18–54 mo). These data must be interpreted in the context that patients with good lung reserve were surgical candidates, and those with respiratory failure and/or poor respiratory reserve at the time of diagnosis of their drug resistance were not surgical candidates.

Drug-related adverse events were common in patients treated for macrolide-resistant MAC lung disease. Most patients received multidrug regimens that included rifabutin, ethambutol, and a parenteral agent, most often streptomycin. These drug toxicity data are pertinent only for drug regimens begun after patients were recognized as having macrolide-resistant MAC lung disease. Significant drug toxicity or adverse events were defined as symptoms or adverse events requiring either dosage change or discontinuation of a drug. The incidence of significant adverse events and the most common adverse events with each drug were as follows: rifabutin, 18 of 44 (41%) patients had gastrointestinal symptoms or polymyalgia/polyarthralgia syndrome; ethambutol, 4 of 40 (10%) patients experienced ocular toxicity; ethionamide, 2 of 7 (29%) patients had gastrointestinal symptoms; streptomycin, 10 of 24 (42%) patients experienced ototoxcity, vestibular toxicity, or hypersensitivity; amikacin, 3 of 11 (27%) patients experienced ototoxicity or vestibular toxicity; ciprofloxacin, 1 of 4 (25%) patients had gastrointestinal symptoms; clofazimine, 2 of 4 (50%) patients had gastrointestinal symptoms; linezolid, 3 of 4 (75%) patients had peripheral neuropathy or anemia; 8-methoxy-fluoroquinolones, 0 of 6; and rifampin, 0 of 15 (most patients receiving rifampin were switched from rifabutin). No patients experienced significant hepatotoxicity with any of the agents used. These figures likely isolates represent an underestimation of the true incidence of significant drug-related adverse events because not all patients were monitored at UTHCT for the duration of macrolide-resistant MAC lung disease therapy.

All isolates were identified as MAC using the commercial DNA/RNA probe (AccuProbe; GenProbe Inc., San Diego, CA). Among the 51 patients identified with macrolide-resistant MAC, all had resistant isolates available for study and 31 had pretreatment or early-treatment macrolide-susceptible isolates available for study. Six of the paired isolates were part of an earlier study that helped define the molecular basis of macrolide resistance in M. intracellulare (29).

All 51 patients had one and most had two or more treatment failure isolates with MICs to clarithromycin of 32 μg/ml or greater. There were 32 patients with pretreatment isolates, with 31 clarithromycin-susceptible isolates having MICs of 8 μg/ml or less (see Table 4).

Initial clarithromycin-susceptible isolates and subsequent resistant isolates were compared by PFGE for 30 patients: 14 patients with nodular disease and 16 patients with cavitary disease. Nineteen pairs of isolates gave indistinguishable patterns, six pairs were probably related (1–3 band difference), two pairs were possibly related (4–6 band difference), and three pairs were unrelated. Four of 14 patients (29%) with nodular disease and bronchiectasis had possibly related or unrelated pairs. Studies of multiple cultures from these patients showed all four to be infected with multiple strains, which produced the observed differences. One of 16 patients (6%) with upper lobe cavitary disease had possibly related pairs and 0 of 16 had unrelated pairs.

A total of 30 preresistant isolates underwent sequencing of the 23S rRNA gene. All had a wild-type adenine at position 2058 and 2059 (Escherichia coli numbering) of the 23S r-RNA gene. Of the 51 macrolide-resistant isolates that were sequenced, 49 (96%) had the expected point mutation of the adenine at position 2058 or 2059 (29–34). Two isolates had two mutations. At position 2058, three had an A→T, 11 had an A→C, and 12 had an A→G. At position 2059, one had an A→T, seven had an A→C, and 17 had an A→G. Overall, there were 26 mutations involving the adenine at position 2058 and 25 mutations involving the adenine at position 2059.

A total of 53 macrolide-resistant MAC isolates from 51 patients underwent hsp65 gene PRA. Forty-one isolates (77%) gave a profile of M. intracellulare, all of which matched the Min v1 profile of Smole and colleagues (18). An additional 12 macrolide-resistant isolates (23%) gave a PRA profile of M. avium, of which eight matched the profile of Mav1 and four matched the profile of Mav2 (18). There were no PRA profiles that matched the described patterns of non–M. avium, non–M. intracellulare MAC species (18). Random isolates (14/53, or 26%) underwent species confirmation using the species-specific DNA/RNA probe (AccuProbe) with 100% concordance with the PRA results. M. avium strains were more common among isolates from nodular disease than among isolates from cavitary disease (9/26, or 35%, vs. 3/27, or 11%; p = 0.04).