Νέες ειδήσεις NTM - Τελευταίες έρευνες

In this retrospective, single-centre, long-term cohort study, the guideline-recommended first-line treatment regimen of rifampicin, ethambutol and a macrolide was better than alternative combination antibiotic regimens at preventing disease progression in Mycobacterium avium complex pulmonary disease (MAC-PD); alternative regimens were not associated with differences in mortality.

The recommended first-line treatment of MAC-PD is rifamycin, ethambutol and macrolides.¹ In clinical practice, one or more of these drugs may be omitted or substituted owing to concerns about adverse reactions or drug–drug interactions.¹ A lack of efficacy outcome data associated with different regimens may be impeding decision-making.¹

Retrospective analysis of patient records (2008–2019) from Osaka Toneyama Medical Centre identified 295 adults aged at least 20 years with a diagnosis of MAC-PD, who started and maintained for at least 6 months either a standard antibiotic regimen (rifamycin, ethambutol, clarithromycin) or an alternative regimen.¹ Patients previously treated with antibiotics for more than 1 month were excluded from the analysis.¹  In total, 238 (80.7%) patients received the standard regimen and 57 (19.3%) an alternative regimen. Analyses were in a 1:1 propensity-matched group of 96 patients (n=48 per treatment regimen) and in the overall population.¹

Compared with patients on the standard regimen, fewer on the alternative regimen had sputum-culture conversion (38/48 [79.2%] vs 29/48 [60.4%]; P<0.05) and there were more treatment failures (16 [33.3%] vs 26 [54.2%]; P<0.05).¹

Susceptibility testing of all 295 patients for resistance to clarithromycin at baseline and during the observation period determined that resistance emerged in 22 patients (9.2%) on the standard regimen and 14 (24.6%) on alternative regimens.¹ Overall, 48 patients died, including 30 from MAC-PD progression. Neither the choice of the standard nor of an alternative regimen was associated with differences in all-cause, or MAC-PD-related mortality.¹

As a retrospective study, limitations included an inability to exclude certain confounding factors, and physician bias regarding timing and selection of antibiotic therapies.¹ The study suggests that patients with MAC-PD receiving a guideline-recommended, combination antibiotic regimen at first line, have the best chance of avoiding disease progression, and alternative regimens should be used only in selected patients or as later-line therapy.¹

Certain patients with Mycobacterium avium complex pulmonary disease (MAC-PD) can experience adverse events (AEs) when receiving clarithromycin as part of combination antibiotic therapy. This retrospective analysis found that switching from clarithromycin to azithromycin may be appropriate in such patients from both efficacy and tolerability perspectives.²

AEs can lead to treatment discontinuation in MAC-PD. Patients with MAC-PD who experience clarithromycin-related AEs may be switched to azithromycin-based combination therapy.² This retrospective analysis identified such patients, evaluated their clinical characteristics, and examined the group’s efficacy and safety outcomes after switching treatment.²  

Records were reviewed of patients with a MAC-PD diagnosis who started clarithromycin-based antibiotic therapy at Kinki-Chuo Chest Medical Center in Southern Osaka between December 2012 and November 2017, then discontinued clarithromycin owing to AEs, and switched to azithromycin-based therapy for more than 12 months.²

Of 365 patients initiating clarithromycin-based treatment during the observation period, 31 switched to azithromycin because of AEs (skin rash, n=13; liver dysfunction, n=6; drug-associated fever, dysgeusia and drug-induced pneumonitis, all n=3; neutropenia, n=2).² Median age was 68 years, most were women (n=20) and most had never smoked (n=23); radiographic patterns were fibrocavitary (n=9), or non-cavitary (n=14), or cavitary (n=8) nodular bronchiectatic disease.²

Median duration of azithromycin treatment was approximately 3.5 years; 96.8% (n = 30 patients) had no AEs, 1 had pruritus.² Three patients switched back to clarithromycin but again experienced AEs (one, dysgeusia; two, skin rash).² Thirteen patients had negative culture conversion, none relapsed during follow-up; all received ethambutol as well as azithromycin as part of their regimen. Seven patients had improved radiological outcomes. Five patients died, one of MAC-PD.²

Limitations of the study include small sample size, the potential for selection bias and a retrospective, single-centre design.² It was also difficult to evaluate causality between clarithromycin and AEs in the context of a multidrug regimen.² Azithromycin-containing regimens seemed to be well tolerated and associated with beneficial bacterial and radiological outcomes in this patient group.² Continuing ethambutol as part of a multidrug regimen may also be important in these patients.² 

Rarity, unspecific presentation and a low index of clinical suspicion can confound timely and accurate diagnosis of non-tuberculous mycobacterial pulmonary disease (NTM-PD). A prediction algorithm, developed using a UK primary care database, offers the possibility of targeted screening for undiagnosed NTM-PD and suggests that prevalence rates may be substantially higher than current estimates³ (see also Ringshausen et al, 2021)⁴.

Although rare, the prevalence of NTM-PD is increasing globally (recent estimates in Europe are 3.3–6.0 cases per 100,000).³ Diagnosis of NTM-PD is challenging because of its similarity to other common respiratory conditions (bronchiectasis, asthma, chronic obstructive pulmonary disease).³ Delays in diagnosis can lead to disease worsening and increased mortality, so it would be advantageous to identify undiagnosed patients at an early stage.³

Machine learning methods were applied to create an algorithm for predicting NTM-PD based on patient medical histories. Patients diagnosed with NTM-PD (September 2003 to September 2017) were identified from diagnosis codes and treatment regimens in a UK primary care database.³ Control individuals with at least one predictor of NTM-PD were also identified within a large sample of individuals from the same database. Pre-diagnosis features of individuals with NTM-PD, and of temporally matched controls, were characterised and used to create the algorithm, which was then validated in a non-overlapping database sample.³

In total, 741 individuals with NTM-PD and 112,874 controls were identified. Individuals with NTM-PD were on average older, more likely to be female, a current or former smoker and to have a lower body mass index than those in the control cohort.³ The annual prevalence of diagnosed NTM-PD in 2016 was 5.1/100,000, increasing to estimates in the range 9–16/100,000 when cases thought likely to have NTM-PD by the algorithm were included.³ The predictive performance of the algorithm was high and improved NTM-PD detection rates by almost 1000-fold relative to random screening. Factors contributing most to algorithm performance were age and the timing of symptoms, of treatments (macrolides and inhaled corticosteroids) and of lung function tests pre-diagnosis.³

The study has limitations; prevalence estimates are based on data that may not represent the wider UK population.³ Fewer primary care practices contributed data to the database in the latter than the earlier part of the study period, and data from settings other than primary care were not included.³ However, the study provides both new epidemiological insights into NTM-PD in the UK and with further validation, the possibility of targeted screening for a rare respiratory disease.³

Diagnosis of non-tuberculous mycobacterial pulmonary disease (NTM-PD) can be delayed by its non-specific presentation and a low index of clinical suspicion owing to its rarity. An algorithm based on historical data from a German health claims database predicted a considerable number of unreported NTM-PD cases nationally. Patient outcomes may improve if such algorithms could be used in clinical practice to facilitate screening and early diagnosis of NTM-PD⁴ (see also Doyle et al, 2020³).

The prevalence of NTM-PD is increasing globally and varies geographically. Diagnosis of NTM-PD is challenging because of symptomatic overlap with underlying conditions and the need for confirmatory clinical, microbiological and radiological data.⁴ Delays in diagnosis can contribute to morbidity and mortality so it would be advantageous to identify undiagnosed patients early.⁴

Algorithms for predicting undiagnosed NTM-PD were developed using machine learning methods.⁴ A sample representing 5% of the German population was created from historical health claims records of more than 4 million individuals.⁴  Within the sample, 218 adult patients diagnosed with NTM-PD (January 2013 to December 2016) were identified by medical claim codes, and 218 adults with no NTM-PD diagnosis were randomly selected as a control group.⁴ Two years of claims data were extracted (pre-diagnosis in the NTM-PD cohort), and variables (age, sex, comorbidities, diagnostic and therapeutic procedures, medications) were compared between the groups. Significantly different variables were evaluated in three different prediction models (classification and regression tree [CART]; boost; random forest), firstly in a training set, then in a validation set to assess performance.⁴ The model with the greatest predictive power was applied to all eligible adults in the population database to identify individuals with claims profiles most resembling those of patients with NTM-PD.⁴

The control group was younger than the NTM-PD cohort (mean, 52.6 vs 61.4 years, respectively).⁴ The most common comorbidities in the NTM-PD group were chronic obstructive pulmonary disease (COPD, 46%), pneumonia (44%), influenza (39%) and asthma (25%). The random forest model had the greatest predictive power.⁴  The most important predictive variables of 23 included were concomitant diagnosis of COPD, a chest X-ray, pneumonia (organism unspecified) and age.⁴ Of 3,196,304 eligible adults in the database in 2016, 488 potentially with undiagnosed NTM-PD were identified. Combining these individuals with those diagnosed increased 2016 prevalence estimates five-fold from 3.79/100,000 to 19.05/100,000.⁴

The model requires prospective clinical validation and has limitations, including the inability to confirm whether those with a high likelihood of NTM-PD had the disease, and the possibility that claims codes among patients with NTM-PD were incorrectly assigned.4 However, the findings imply large numbers of unreported NTM-PD cases in Germany that could be detected at an early stage by implementation of a suitably validated prediction algorithm.⁴

“These fascinating studies utilise machine learning to identify patients in healthcare databases that may have NTM-pulmonary disease. If the predictive algorithms are validated, it would represent a major step forward in our ability to identify affected individuals at an earlier stage in their disease course and thereby offer an opportunity to prevent disease progression.” Charles Haworth, Papworth Hospital, UK