A better understanding of the epidemiology of mycobacterial infections leads the research agenda issued by national NTM guidelines [1]. NTM infections are usually not reported to public health departments, thus data on incidence and species distribution are sparse. However there have been suggestions that NTM disease is increasing in some areas [2, 3], making the demography and species distribution of these infections important to understand. The mid-Atlantic and southeastern United States are known to be high incidence regions for mycobacteria, likely related to a high prevalence of NTM in local soils [4]. The University of Virginia hospital is a tertiary referral center that receives patients from across the region. Thirty years have passed since the last systematic report of mycobacterial infections at our institution [5]. Thus we sought to describe the epidemiology, clinical characteristics, and radiology of all mycobacterial infections seen at our hospital from 2001-2009. To our knowledge, this is among the largest radiographic series of pulmonary mycobacterial infections reported, with 244 chest CT scans available for analysis.

The clinical approach to mycobacterial infection is vexing until speciation is confirmed, a process that can take weeks. In particular, when faced with positive acid-fast bacilli smears or positive growth, the dilemma emerges of whether to treat empirically for TB or an NTM or wait. Rapid nucleic acid amplification tests for Tb can be useful in this scenario but are only approved for sputum or bronchoalveolar lavage material and are not always available. The decision to defer treatment for TB risks ongoing infectiousness and clinical deterioration, while presumptive treatment for TB incurs significant polypharmacy and may trigger contact investigations. We therefore sought to identify clinical parameters that could aid in predicting TB infection over other NTM in our setting.

This current study of mycobacterial infections is drawn from a geographic area prevalent for both NTM and TB and spans nearly a decade. The study describes important epidemiology regarding demography, speciation, and pathogenesis, with particular attention paid to pulmonary mycobacterial infections and chest CT scan findings. Given the lack of mandatory reporting of NTM, studies from large referral institutions such as this are critical to understand regional patterns or emerging trends. We observed a non-statistically significant rise in the number of mycobacterial isolates over the last decade, significant species diversity, with more total cases of pulmonary NTM than during an equivalent time period from our institution thirty years ago [5]. Estimation of the true change in incidence is difficult from these data given alterations in population referral, potentially heightened clinician suspicion of NTM, and improved diagnostics, however our findings complement trends in other parts of the world that suggest increasing rates of NTM disease and species diversity [2, 3, 7].

Among patients meeting criteria for NTM lung disease, MAC remained the overwhelming pathogenic species, consistent with previous findings at our institution and typical of most NTM epidemiologic reports [8]. After MAC we observed significant diversity of 8 other species, as well as 8 mixed infections, which suggests that clinicians should not readily discount a particular species in this region. We looked for risk factors for particular NTM lung species, including age, gender, chronic obstructive pulmonary disease, cystic fibrosis, other lung diseases, diabetes, end stage renal disease, collagen vascular disease, cancer, HIV and transplantation. The only obvious host characteristic that was linked with a particular NTM lung disease was male gender with M. xenopi and cystic fibrosis with M. abscessus. The latter association has been appreciated before [9] but not usually to this extent, where fully 38% of our M. abscessus infections were associated with CF. This may be a feature of our center which is a tertiary referral center for lung transplantation. Of course we may have missed identifying risk factors due to low power for most species besides MAC, but it still suggests to us that in our region mycobacteria do not present stereotypically, thus pathogenesis is likely not species-specific, and no particular species can be easily considered low virulence (as M. fortuitum is sometimes considered [1]). Another finding was the relatively low proportion of individuals that strictly met ATS/IDSA criteria for NTM lung disease: 82 of 342 pulmonary NTM infections or 24%. This contrasts with other studies where more than half of all pulmonary NTM patients met these criteria [10]. Our suspicion is that this does not reflect true differences but that our proportion of patients that met criteria for NTM lung disease was an underestimate because we relied tightly on chest CT scans performed at our institution within 6 weeks of the isolate.

Over two thirds of patients with a pulmonary mycobacterial isolate had a CT scan available for analysis, making this one of the largest radiographic series of NTM lung infections available. NTM lung disease has historically been ascribed two radiographic categories: fibrocavitary disease similar to Tb, especially in males, or nodular bronchiectatic disease, especially in females. We could not discern such a distinction and found significant overlap across the species in terms of cavities, nodules, and bronchiectasis regardless of gender. Previous reports have also failed to find a difference in radiologic appearance of NTM infection based on the mycobacterial species [11–13]. However a few trends were revealed. Notably, M. fortuitum and M. kansasii had similar patterns, with a larger proportion with cavities than the most common NTM species, MAC. Cavitation has been thought to be rare with M. fortuitum and the other rapidly growing mycobacteria [14]. Among the patients with M. fortuitum and cavities, all had an immunosuppressive condition or pre-existing lung disease that may have contributed. The high frequency of cavities is well known for M. kansasii [15, 16], often times with thinner walled cavities that Tb, however this latter feature was not reliably described in our radiology reports. Cavities have been reported in up to 50% CT scans in patients with active MAC infection [17], yet others report more bronchiectasis and nodules [13, 18, 19]. Our series found the latter, with nearly two-thirds of our 86 patients with pulmonary MAC infection and CT scans exhibiting nodules alone or nodules with bronchiectasis. The strength of our series is that it was larger than any of these previous studies.

We acknowledge that our radiologic analysis has limitations. Radiologists may have received information on the microbiologic diagnosis from the clinicians, and this lack of blinding may have led to bias in reporting findings historically associated with the known diagnosis or present on prior scans that we did not review. Furthermore, it may be that patients with characteristic findings on chest radiograph were not sent for chest CT, as may be evidenced by the 21 patients with TB from pulmonary specimens, of whom only 13 had a chest CT available for analysis. On the other hand these limitations are the norm in clinical practice.

The first scenario clinicians face is a positive acid-fast smear or mycobacterial culture and determining a management course while awaiting results of speciation. In comparing risk factors for TB or NTM among patients with pulmonary mycobacterial specimens, likelihood ratio analysis of immigrant status led to post-test probabilities of TB exceeding 80%. The magnitude of this observation may be secondary to reporting bias since a clinician may have been more likely to document in the chart a patient as an "immigrant" if TB was confirmed. Furthermore, the term "immigrant" was not defined or validated. Therefore although our rates of immigrants among Tb and NTM patients may be imperfect they are likely reasonable. Indeed, our proportion of Tb seen among immigrants, 67%, is almost identical to our state's rate of Tb among foreign-born individuals. On the topic of specimen type, our data would emphasize the importance of sputum over bronchoscopy to evaluate for pulmonary Tb, since sputum was as or more sensitive than bronchoalveolar lavage fluid for Tb growth in all 9 instances where both were sent.

Finally, our series revealed a large number of "other" NTM, including at least 3 infections of M. peregrinum, M. terrae, and M. xenopi as well as more than a dozen others. This adds important detail to the known epidemiology of NTM. For instance, M. simiae has been appreciated mostly from the southwestern United States [1], but not our region. Additionally, the NTM guidelines describe M. terrae as "potentially pathogenic," which we would strongly support. And M. xenopi is reported as rare in the US, although we observed 5 infections with some extensive disease.

In summary, we found a substantial number and diversity of mycobacteria isolated from patients at our center throughout the past decade. Series such as this from other centers are welcomed to better inform the complex and emerging epidemiology of mycobacterial infections.