Predictors of developing Mycobacterium kansasii pulmonary disease within 1 year among patients with single isolation in multiple sputum samples: A retrospective, longitudinal, multicentre study

The clinical significance of a single Mycobacterium kansasii (MK) isolation in multiple sputum samples remains unknown. We conducted this study to evaluate the outcome and predictors of developing MK-pulmonary disease (PD) within 1 year among these patients. Patients with a single MK isolation from ≥3 sputum samples collected within 3 months and ≥2 follow-up sputum samples and chest radiography in the subsequent 9 months between 2008 and 2016 were included. The primary outcome was development of MK-PD within 1 year, with its predictors explored using multivariate logistic regression analysis. A total of 83 cases of a single MK isolation were identified. The mean age was 68.9 ± 17.9, with a male/female ratio of 1.96. Within 1 year, 16 (19%) cases progressed to MK-PD; risk factors included high acid-fast smear (AFS) grade (≥3), elementary occupation workers, and initial radiographic score >6, whereas coexistence with other nontuberculous mycobacterium species was protective. Among patients who developed MK-PD, all experienced radiographic progression, and 44% died within 1 year. Although a single MK isolation does not fulfil the diagnostic criteria of MK-PD, this disease may develop if having above-mentioned risk factors. Early anti-MK treatment should be considered for high-risk patients.

SCIeNTIfIC REPORtS | (2018) 8:17826 | DOI: 10.1038/s41598-018-36255-w NTM species is required to fulfil the microbiological criteria from current guideline 1 . Therefore, a single NTM isolation from multiple respiratory specimens is usually considered colonization or contamination rather than indicative of a true pathogen. In a study conducted in South Korea, 26 (14%) of 190 patients with a single sputum isolation of pathogenic NTM met the diagnostic criteria for NTM-PD within a median follow-up period of 16 months, and none of the 8 patients with a single MK isolation had a subsequent positive culture 14 . Under this clinical entity, follow-up sputum sampling and clinical monitoring are necessary to diagnose NTM-PD; yet, the clinical outcome and optimal duration of follow-up remain uncertain 14,15 . Furthermore, patients may have different NTM species in the respiratory tract presenting as transitional, alternating, or simultaneous pattern in chronological order 1,16,17 . Little is known of the clinical significance of coexistence of different NTM species. In Taiwan, MK is the third most common NTM causing PD, and the numbers of MK isolates increased 4.7-fold from 2010 to 2014 9 . Being familiar with the outcome of different clinical entities of MK is crucial in clinical practice. We therefore conducted this retrospective, longitudinal cohort study to investigate the predictors of developing MK-PD within 1 year among patients with single MK isolation from multiple sputum samples, with a special emphasis on coexistence of NTM species other than MK.

Results
Study population. Figure 1 shows the flowchart of patient selection and the enrolment criteria. Between 2008 and 2016, a total of 1,852 respiratory MK isolates in 1,183 patients were identified from six hospitals. By applying the selection criteria, a total of 83 (7.0%) cases of single MK isolation were finally selected for further analysis.
Among the 83 patients, the mean age of the patients was 68.9 ± 17.9 years, with a male/female ratio of 1.96. Among the 83 patients, 52% were ex-smokers and 54% had received education for <6 years. 20 (24%) cases were employed in elementary occupations. Of them, 18 were in industrialized areas, including 14 construction laborers and 4 iron-steel manufacturing laborers. Of the remaining two patients, one was a truck driver and the other was a cleaner. The most common pulmonary comorbidity was chronic obstructive pulmonary disease (COPD: 35%), and the most common systemic comorbidity was chronic kidney disease stages 3-5 (23%). Higher (but nonsignificantly) prevalence rates of COPD and pneumoconiosis (respectively, 50% vs. 28%, p = 0.086; 11% vs. 0%, p = 0.060) were   There was no significant difference in clinical characteristics between the 15 patients with coexistence of other  NTM species and the other 68 without such coexistence (Table 1), except that the coexistence group had a significantly higher prevalence of congestive heart failure (33% vs. 7%, p = 0.018). The initial symptoms, laboratory data, lung function, radiographic findings, and sputum mycobacteriologic results were similar between the two groups.
Outcome and risk factors of MK-PD. Within 1 year, MK-PD developed in 16 (19%) patients, including 6 (38%) in the second quarter and 8 (50%) in the third quarter (Fig. 2). All of these patients experienced radiographic progression during follow-up. The radiographic findings during MK-PD development were FC pattern in 10 (63%) patients and NB pattern in 6 (38%), and all showed multifocal involvement.
Multivariate logistic regression analysis was performed to investigate the independent predictors for developing MK-PD. We found a statistical separation phenomenon existed in two variables, coexistence with other NTM species and high-grade positivity in sputum AFS. One-year outcome. Treatment courses and 1-year mortality are summarized in Table 3. Treatment for MK was more frequently, though not significantly, prescribed in the 16 cases developing into MK-PD according to either intention-to-treat analysis (29% vs. 13%, p = 0.087) or per-protocol analysis (13% vs. 3%, p = 0.166). Of the 16 patients who developed MK-PD within 1 year, 7 (44%) died, 4 of whom died of MK-PD. By contrast, 6 (9%) of the 67 without development of MK-PD died (p = 0.002). Neither of the two patients with MK-PD who received standard anti-MK treatment died, whereas two patients who received transient anti-MK treatment died.

Discussion
To our knowledge, this is the first longitudinal, multicentre study investigating the incidence and predictors of developing MK-PD within 1 year among patients with single MK isolation from multiple sputum samples. There were two major findings in this study. First, MK-PD developed in 19% of cases, with 88% of the MK-PD occurring in the second and third quarters after the index date. All patients who progressed to MK-PD had typical radiographic patterns with multifocal involvement, and radiographic progression occurred thereafter. The 1-year mortality rate was 44% in MK-PD patients with MK-PD, which was 4.9 times higher than in those without MK-PD. Second, high-grade sputum AFS at initial presentation, elementary occupations, and higher initial radiographic scores (>6) were independent risk factors for developing MK-PD within 1 year, whereas coexistence with other NTM species was protective.
Though a single respiratory isolation of pathogenic NTM, such as MAC, M. abscessus, and MK, is insufficient to establish the diagnosis of NTM-PD 1 , the results of previous studies suggest that it is clinically significant in appropriate clinical settings, especially in cavitary lung disease 5,[17][18][19] . Some experts also suggest a lower diagnostic threshold for patients with positive respiratory cultures of MK, particularly in people living with human immunodeficiency virus 18,20,21 . Another study obtained the opposite result, showing that 14% of patients with a single sputum isolation of NTM eventually developed NTM-PD within a median follow-up period of 16 months 14 . Another retrospective observational study had a similar finding, showing that none of 18 patients with a single MK isolation developed MK-PD within a median of 12 months 22 . However, it's difficult to draw a definite conclusion due to the limitation of power (small sample size). Therefore, an appropriate follow-up period is proposed to be essential to determine the clinical relevance of a single MK respiratory isolation 14,15 . The present study demonstrated that 19% patients with a single respiratory isolation of MK would progress to MK-PD, and they should be followed for at least 9 months to determine its clinical relevance. Given the high probability of adverse drug reactions, regular monitoring is recommended, rather than immediate anti-MK treatment.
Among patients with a single MK respiratory isolation who developed MK-PD within 1 year, it is striking that all experienced radiographic progression and 44% of them died. The mortality rate in this study is much higher than that in other reports, ranging from 9% to 15.8% 3,23 , probably because the present study was conducted in medical centres where patients tended to have high disease severity. In addition, the diagnosis of MK-PD in the present study strictly followed the American Thoracic Society/Infectious Diseases Society of America guidelines 1 , whereas previous reports exclusively employed the microbiological component of the current guidelines, which may underestimate the severity of disease 9 . Given the poor outcome, predicting subsequent progression  16,19 . This highlights the dynamic nature of NTM and puzzles clinicians. However, the clinical significance of this phenomenon is unclear. In patients receiving treatment for MAC-PD, single respiratory isolation of M. abscessus probably requires no therapy 17 . However, the change from MAC to M. abscessus is usually accompanied by symptomatic and radiographic worsening 24 . Unlike other studies, the present analysis revealed that coexistence of MK and another NTM may be associated with a lower risk of MK-PD and initial high-grade sputum AFS is a risk factor for MK-PD, supporting the hypothesis that quantitative organism load might determine the clinical relevance of NTM-PD 8,11,25 . Coexistence with other NTM species may therefore reflect that MK is not the dominant microorganism.
The epidemiology of MK is predominantly urban and has been associated with high-density and low-income communities 5,7 . The major reservoir has been postulated to include water systems associated with habitation or industry, and infection probably occurs via an aerosol route 1,11,20 . These factors may explain why the incidence of MK-PD was higher among the patients employed in elementary occupations in the present study; such workers extensively use aerosolized water for dust control 5,20,26 .   There are several limitations of the present study. First, the lack of subtyping for MK precluded us from distinguishing its pathogenic (such as MK subtype 1) and nonpathogenic subtypes 27 . Second, we excluded patients providing less than three sputum samples even if single MK isolation was noted, which may have resulted in underestimation of the incidence of subsequent MK-PD. Third, no standardized microbiological or radiographic follow-up protocols were used in this retrospective study. Fourth, the data were retrieved from medical centres in Taiwan and may not be generalizable to all populations.
In conclusion, approximately one-fifth of the patients with single MK isolation from multiple sputum samples progress to MK-PD within 1 year; the majority of progressions occur in the second and third quarters thereafter. However, once MK-PD develops, radiographic progression is inevitable, with 1-year mortality of >40%. Early treatment for MK should be considered for vulnerable populations. Having an elementary occupation, high-grade sputum AFS positivity, and high initial radiographic score (>6 points) are risk factors for MK-PD, whereas coexistence with NTM other than MK is protective.

Methods
Study population. This retrospective study was conducted in two medical centres, the National Taiwan University Hospital (NTUH) and Kaohsiung Medical University Hospital (KMUH), and their four branch hospitals. This multicentre study was approved by the medical centres' institutional review boards (NTUH REC 201508017RIND and KMUH IRB-SV[I]-2015200266) and the need for informed consent was waived because data utilized in this retrospective study have been de-identified.
From January 2008 to December 2016, respiratory specimens were retrieved from the mycobacteriology databases. Mycobacteriologic examinations were performed as described previously 9 . Only patients with new episodes of MK since 2008 were selected 9 . Patients who provided ≥3 sputum samples within 3 months with only one MK isolation, and who had ≥2 follow-up sputum samples and chest radiography in the subsequent 9 months were selected into the study. Only the first episode of single MK isolation for each patient was selected for analysis. We excluded patients lacking demographic data and having TB concomitantly. The index date was defined as the date when the index culture of MK isolation was plated. The patients were followed up until diagnosis of MK-PD, death, or 1 year after the index date. Chest radiographs and computed tomographic scans were interpreted independently by two pulmonologists. We categorized the patterns as fibrocavitary (FC) and nodular bronchiectatic (NB) and the extent as focal and multifocal. Radiographic scores for severity assessment were recorded as previously described 29 .
The MK treatment administered was analysed in two ways: (1) intention-to-treat analysis for patients who had ever received any drugs against MK, regardless of duration; and (2) per-protocol analysis for patients who had received combination chemotherapy against MK for >2 months in the first 3 months after the index date.
Outcome assessment. The primary outcome was development of MK-PD within 1 year of the index date.
The diagnosis of MK-PD was according to current guidelines 1 . The secondary outcome was radiographic progression and mortality within 1 year for those developing MK-PD. MK was considered the cause of death if no pathogens other than MK were identified and radiographic progression of MK-PD was noted.
Statistical analysis. Continuous variables are presented as mean ± standard deviation or median with interquartile range and were compared using independent samples t tests. Categorical variables are expressed as percentages and were compared using the chi-squared test or Fisher's exact test, as appropriate. The independent factors associated with development of MK-PD were determined using multivariate logistic regression analysis. Time-to-event curves for development of MK-PD were generated and compared using the log-rank test. Statistical significance was set at p < 0.05 (two-sided). All statistical analyses were performed using IBM SPSS version 22.0 (IBM, Armonk, NY, USA).

Data Sharing Statement
All data were deposited in the Information Technology Office of National Taiwan University Hospital and Statistical Analysis Laboratory, Department of Medical Research, Kaohsiung Medical University Hospital. The data were not available for sharing without permission.

Cause of Death
Sepsis with bacterial pathogen 3 (19%) 4 (6%) Mycobacterium kansasii 4 (25%) 0 (0%) Others 0 2 (3%) d Table 3. Treatment and 1-year outcome of patients, stratified by progression to Mycobacterium kansasii pulmonary disease (MK-PD). Data are number (percentage) or median ± interquartile range. p value was calculated using the chi-squared test unless otherwise mentioned. a Patients who had ever received any drugs against MK, regardless of duration. b Patients who had ever received combination chemotherapy against MK for more than 2 months in the first 3 months. c p value was calculated by log-rank test. d The cause of death was lung cancer in one and acute myocardial infarction in the other.