Citations

Abstract RATIONALE: Research since 2017 has highlighted artificial intelligence (AI)-based technologies to assist in detecting and/or classifying lung nodules on Computed Tomography (CT) scans. However, few have combined detection and classification tasks, and most have required extensive hands-on image manipulation to process the cases, which is both costly and time-consuming. Here, we present a closed loop, fully automated software package, Bronchosolve, which handles cases from scan to result without manual intervention. METHODS: The software package combined pre-processing, analysis, and result generation system including a deep learning convolutional neural network (CNN) for assessment of suspicious pulmonary nodules. Model inputs were full CT cases formatted in Digital Imaging and Communications in Medicine (DICOM), without demographics or clinical variables. Automated steps included 1) automatic selection of the optimal CT series for analysis; 2) normalization and pre-processing of the series; 3) AI-based analysis of the images to both detect and classify suspicious lung lesions; and 4) generation of analysis report. The system was trained using a multi-center dataset of over 2,358 cases of suspicious lung nodules, including cancers and non-cancers. A United States-based, multi-site cohort (n=184) was used for initial algorithm validation where positive cases were defined as biopsy-proven cancer within 1 year, and negatives as negative biopsy or 2+ year negative follow-up. RESULTS: In the validation set, the system achieved a technical completion rate of 100%, automatically for all cases. Median patient age was 62.5 years (interquartile range 58.5 to 66.5 years), 45% were former smokers, 55% were current smokers, and 40% were female. The AI cancer risk (CA-risk) model achieved an area under the receiver-operating-characteristic curve (AUC) of 0.828 [CI: 0.773 – 0.883] in differentiating cancer from non-cancer in cases with known suspicious pulmonary nodules, as confirmed by biopsy, surgery, and/or follow-up. These results were statistically superior to Lung-RADS pAUC of 0.669 [CI: 0.601 – 0.737] and trended better than Brock model AUC of 0.783 [CI: 0.723 – 0.842], Figure 1. At the Youden operating point, model sensitivity was 73.1% [CI: 66.7 – 79.5%], and specificity was 82.5% [CI: 77.0 – 88.0%]. The model's performance was consistent across various CT scanner manufacturers and tested slice thickness. CONCLUSIONS: The closed loop, automated software system, Bronchosolve, consistently processed and analyzed a wide range of CT scans, providing accurate assessment in risk-classifying suspicious lung nodules. This approach has the potential to improve non-invasive assessment of lung nodules eliminating the need for additional user training and time.

Abstract Chronic obstructive pulmonary disease (COPD) and pulmonary emphysema were the third-leading cause of death globally in 2019. Besides alpha-1 antitrypsin deficiency, emphysema lacks disease-modifying medications due to its heterogeneity and there remains a need to further investigate emphysema subtypes to advance its diagnosis and treatment. Recent unsupervised machine learning of the texture and anatomical location of emphysema on computed tomography (CT) images of the lung defined six quantitative CT emphysema subtypes uniquely associated with specific gene variants, environmental exposures, symptoms, physiology and clinical events. We aim to identify associations of plasma proteins with each CT emphysema subtype cross-sectionally in a multi-ethnic general population sample. MESA is a prospective cohort study that recruited 6,814 participants from six communities who self-identified as White, Black, Hispanic, or Asian race/ethnicity, ages 45-84 years old and free of clinical cardiovascular disease in 2000-02. The MESA Lung Study performed CT scans among 3,137 participants in 2010-12, which were labeled with previously described (PMID: 37268414) CT emphysema subtypes. Proteomics was assessed using SomaScan version 4.1 assay (7,596 plasma proteins). Linear regression models were adjusted for age, height, weight, sex, race, educational attainment, smoking status, pack years, urine cotinine level, body-mass index (BMI), estimated glomerular filtration rate (eGFR), CT scanner manufacturer and other CT emphysema subtypes. Results were adjusted to reflect a False Discovery Rate (FDR) p-value of less than 5% using the Benjamini-Hochberg method. Among 2,343 participants with both CT and proteomic data (mean age 69.5 years, 52.3% women, 23.7% with COPD), we identified statistically significant associations of proteins with the vanishing lung subtype (42 proteins; top hit Natural cytotoxicity triggering receptor 3 ligand 1, q = 2.9[asterisk]10-9), restrictive combined pulmonary fibrosis/emphysema (CPFE) subtype (11 proteins; top hit Integrin alpha-2/b1, q=1.0[asterisk]10-5), the combined bronchitis-apical emphysema subtype (9 proteins; top hit Desmoglein-2, q=8.8[asterisk]10-13), and the diffuse emphysema subtype (4 proteins; top hit Pulmonary surfactant-associated protein D, q=5.7[asterisk]10-3). No significant proteins were associated with the obstructive CPFE or senile subtypes after FDR correction. Most proteins were uniquely associated with one CT emphysema subtype, while a few proteins were associated with more than one subtype. This study identified specific proteins for different CT emphysema subtypes, many of which have been related less specifically to emphysema, including pulmonary surfactant production, anti-inflammatory pathways, and mitochondrial deregulation, amongst others. Validation of these proteomic associations can provide insight into further pathway assessment, clinical biomarker discovery, and targeted prevention of emphysema in the general population.

Abstract RATIONALE: Airway-to-lung ratio (AirLR), calculated as the mean of airway lumen diameters at standard anatomic locations divided by the cube-root of total lung volume, is associated with COPD risk regardless of smoking status, and we have previously shown an association between total-pulmonary-vascular-volume-to-lung-size relationship with AirLR. [PMCID: PMC11284327] Using dual energy computed tomography (DECT), we recently developed a deep learning-based algorithm for pulmonary arterial segmentation to compute pulmonary-arterial-to-lung ratio (ArtLR) which correlated significantly with AirLR and markers COPD. Here we automate image processing in a larger cohort to estimate ArtLR, introduce an alternative ArtLR standardization and correlate it to COPD-related measures of lung structure and function. METHODS: Using a SPIROMICS sub-cohort (110 participants), imaged via contrast enhanced DECT at functional residual capacity (FRC), an automatic deep learning pipeline segmented the arteries from the pulmonary vascular tree. Another pipeline automatically extracted the centerlines and calculated average arterial segment diameters. A paired non-contrast total lung capacity (TLC) scan was registered to the DECT FRC scan to transfer airway segment labels from the TLC to FRC domain. AirLR was calculated as the mean of airway lumen diameters at standard anatomic locations (trachea-to-subsegments) divided by cube root of lung volume [TLC or FRC]. A simple Long Short-Term Memory (LSTM) network was used to find corresponding anatomic arterial segments. Associations between ArtLR and %Emphysema, AirLR, pre- and post-bronchodilator FEV1/FVC, and a texture-based (AMFM) measure of %broncho-vascular bundles was assessed using linear regression to adjust for age, sex and BMI. RESULTS: In adjusted analyses, a 1-SD decrement in ArtLR-TLC was associated with 4.198 units increase in %Emphysema-910 (95%CI: 1.360-7.037; p = 0.0041), and 0.8198 units decrease in %Bronchovascular (95%CI: 0.5037-1.136; p < 0.0001). A 1-SD increment in ArtLR-TLC was associated with 0.2515 SD increase in AirLR-Outer (95%CI: 0.04947SD-0.4536SD; p = 0.0152), and 0.02083 units increase in PreBronch-FEV1FVC (95%CI: 0.001814-0.03984; p = 0.0321). Sex and BMI show significant associations with ArtLR-TLC, where female is associated with 1.068 SD decrease in ArtLR, and a unit increase in BMI is associated with 0.05788 SD increase in ArtLR (both p-values < 0.0001), and age was not association with ArtLR. CONCLUSIONS: An automated pipeline for pulmonary arterial tree segmentation and standardization from contrast enhanced DECT FRC scans demonstrates a significant relationship between ArtLR and AirLR as well as between ArtLR and %emphysema, and FEV1/FVC. A standardized method of arterial tree size assessment provides for expanded exploration of vascular vs airway disease etiologies.

Abstract Background. Sarcopenia (i.e. skeletal muscle loss) and physical frailty are significant contributors to adverse clinical outcomes in lung cancer patients. Nutritional intake and physical activity influence skeletal muscle mass and strength. We determined the impact of lung cancer on clinical outcomes in hospitalized patients with sarcopenia, frailty, or sarcopenia+frailty. Design. The Nationwide Inpatient Sample (NIS) (years 2018-2020) from the United States was analyzed to study adult hospitalized patients with a diagnosis of lung cancer. Sarcopenia and frailty were defined by ICD10 codes that have been previously validated. Multivariate regression models were generated to determine the clinical impact of sarcopenia, frailty, or sarcopenia+frailty in patients with a diagnosis of lung cancer as compared to a general medical population (GMP) without lung cancer. Age was restricted to >40 years old for both cohorts because lung cancer is rare in those less than 40 years old. Models were adjusted for age, race, sex, and comorbidities (cardiovascular disease, anemia, coagulopathy, HIV, depression, diabetes mellitus, and COPD). Results. Compared to the GMP (n=126,099), patients with lung cancer (n=241,913) were older (mean+SD age: 69.6+10.0 vs. 67.6+13.3 years, p<0.001), and had a higher prevalence of sarcopenia (22.5% vs 9.0%, p<0.001), frailty (53.4% versus 49.8%, p<0.001), and sarcopenia+frailty (15.3% versus 6.8%, p<0.001). COPD was common and found in 52.3% of lung cancer patients and 20.2% of the GMP. In lung cancer patients, the most common medical reasons for admission were sepsis, pneumonia, and COPD exacerbation. In GMP patients, the most common medical reasons for admission were sepsis, congestive heart failure, and acute kidney injury. In lung cancer patients, frailty had the greatest association with mortality (unadj OR 3.12, 95% CI: 3.02-3.23; adj OR 3.04, 95% CI: 2.91-3.17). Sarcopenia+frailty was most associated with greater healthcare utilization including greater healthcare cost (unadj OR 1.31, 95% CI: 1.29-1.32; adj OR 1.28, 95% CI: 1.26-1.30) and length of stay (unadj OR 1.57, 95% CI: 1.55-1.58; adj OR 1.50, 95% CI: 1.48-1.52). Sarcopenia+frailty were more likely to require additional healthcare facilities upon discharge and were the least likely to be discharged to home (unadj OR 0.32, 95% CI: 0.32-0.33; adj OR 0.35, 95% CI: 0.33-0.36). (see Table 1). Conclusion. While frailty was most associated with in-hospital mortality in lung cancer patients, sarcopenia+frailty was most associated with greater healthcare utilization. Our findings highlight the need to develop targeted interventions to improve clinical outcomes in lung cancer patients with sarcopenia and/or frailty.

Abstract RATIONALE: Vascular dysfunction contributes to pathological tissue remodeling in aging-associated disease such as Idiopathic Pulmonary Fibrosis (IPF). IPF is an incurable lung disease characterized by progressive accumulation of scar-forming fibroblasts and several vascular alterations, including endothelial transcriptional changes leading to a pro-fibrotic endothelial secretory phenotype. To date, the mechanisms governing these transcriptional changes remain largely unexplored. To fill this gap, we sought to investigate whether microRNAs (miRNAs), small non-coding RNA that function as key gene expression regulators, modulate the transcriptional changes of the endothelium associated with progressive pulmonary fibrosis. METHODS AND RESULTS: miRNA expression profiling of lung ECs (LECs) sorted from aged mice with persistent lung fibrosis revealed upregulation of miR-205-5p. Further analysis demonstrated that endothelial miR-205-5p was strongly upregulated in an additional mouse model of progressive lung fibrosis induced by repetitive bleomycin dose as well as in human LECs purified from IPF lungs. RNA-seq analysis of normal human LECs overexpressing miR-205-5p revealed reduction of genes associated to endothelial identity (e.g. PECAM1, ERG, NOS3), and an increase of genes encoding for soluble pro-inflammatory and pro-fibrotic molecules, including TGFBI, TGFB2, TGFB3, CXCL10, CCL5, IL1B and FAS compared to normal LECs transfected with a control mimic. Consistently, conditioned medium from miR-205-5p overxpressing LECs exerted a pro-fibrotic activity on quiescent human lung fibroblasts, as demonstrated by upregulation of a panel of profibrotic transcripts (e.g. COL1A1, CTHRC1, FN1 and TNC) and elevated Collagen I deposition compared to human lung fibroblasts exposed to conditioned medium generated by human LECs transfected with a control mimic. CONCLUSION: Altogether, our findings suggest that lung endothelail miR-205-5p induces loss of endothelial cell identity and drives the secretion of pro-fibrotic molecules that promote the transition of quiescent fibroblasts into scar-forming fibroblast through paracrine signal. Our data set the stage for further investigations into the role of miR-205-5p and associated pathways in lung vascular injury responses, with the long term goal of targeting endothelial dysfunction for slowing the progression of pulmonary fibrosis.

Abstract Rationale: Heart failure (HF) is often characterized by elevated left ventricular filling pressures (LVFP) and pulmonary congestion. However, LVFP may not always predict congestion. Pulmonary congestion can be quantified by computed tomography (CT) imaging. We hypothesized that extravascular lung water quantified by CT may differ in HF groups such as HF with preserved ejection fraction (HFpEF) vs HF with reduced ejection fraction (HFrEF), inpatients with acute decompensated heart failure (ADHF) vs stable and this would be related to echocardiography (echo) derived estimates of LVFP. Methods: We enrolled and studied 83 patients with HF who were either admitted or had a clinic visit at Mayo Clinic, Rochester, MN from May 2018 through March 2019. All subjects underwent comprehensive echo and supine thoracic CT scans. CT lung fluid was quantified and reported as proportion (%) of total lung volume. CT attenuation in lung tissue was also measured. LVFP was evaluated according to standardized echo diagnostic algorithms. Patients were stratified based on clinical diagnosis of HF as inpatients with HFpEF (n=26, 75.6±8.5yrs, M=50%, BMI=36.2±8.9kg/m2) inpatients with HFrEF (n=33, 75.6±14yrs, M=79%, BMI=31.1±9.1kg/m2), inpatients with ADHF (n=14, 72.6±14.5yrs, m=64.2%, BMI= 32.2±7.6Kg/m2) and stable outpatient HF controls (n=10, 65±13 yrs, m=90%, BMI= 29.5±5.2Kg/m2). All four groups were compared using one-way ANOVA followed by a post-hoc analysis. A correlation between lung fluid % and LVFP was performed using Spearman's correlation test. Results: When HF groups were compared according to CT % lung fluid, significant differences were found between HF groups (p=0.004). Post-hoc analysis revealed a significant difference in % lung fluid as HFrEF group had higher mean value in % lung fluid compared to HF control group (p=0.004). Similarly, ADHF group had higher mean % lung fluid compared to HF control group (p= 0.007). However, we found no correlation between LVFP and % lung fluid. The rightward shift in CT attenuation for HFpEF, HFrEF, and ADHF subjects, as displayed in figure 1, suggests greater levels of % lung water in these subjects compared to controls. Conclusion: These data suggest that CT measures of lung water differ in subgroups of HF patients and may not correlate with LVFP.

Abstract Food allergy reactions emerging after organ transplantation, particularly following lung transplantation, are uncommon but pose significant clinical challenges. While most documented cases have involved liver transplants and pediatric patients, instances in lung transplant recipients remain rare. Here, we present a case of acute bronchospasm triggered by hummus ingestion, requiring mechanical ventilation in a post-lung transplant patient. This case underscores the need for heightened awareness of potential allergenic reactions in transplant recipients and emphasizes the importance of prompt recognition and management of such reactions in the critical care setting. A 61 year old female experienced an acute respiratory event requiring emergent intubation shortly after ingesting hummus. Initial treatment included duonebs and bronchoscopy, revealing a normal airway architecture but notably stiff lungs. She responded to epinephrine administration. Post-event serum tryptase was within normal limits, suggesting the reaction was primarily respiratory in nature. Despite having a known peanut allergy in her organ donor, her own history was negative for peanut IgE reactivity. Subsequently, she underwent extensive blood IgE testing, confirming negative IgE to peanut, yet yielding positive or equivocal results to chickpea and sesame. Skin testing was positive for “Sabra Snack Hummus,” suggesting reactivity. Prick-to-prick skin testing with chickpea, however, was negative. Additionally, IgE component testing identified Ara h2, a peanut protein with potential cross-reactivity to other legumes, including chickpeas. Managing peanut allergy in post-transplant patients requires careful consideration to prevent severe allergic reactions. Screening is crucial when a donor has a known food allergy, as this can help identify potential allergic responses due to transferred donor immune cells. If a peanut allergy arises post-transplant, strict avoidance of peanuts and cross-reactive allergens, such as certain nuts and legumes, is advised. Transplant recipients with confirmed peanut allergies should carry an epinephrine auto-injector to immediately treat potential anaphylaxis. The pathophysiology of post-transplant food allergies remains under investigation. One theory suggests that donor-derived B lymphocytes producing peanut-specific IgE may enter the recipient's circulation, leading to sensitization. Alternatively, IgE-sensitized mast cells and basophils within the transplanted tissue may react directly in the recipient's lungs or lymph nodes. Another possibility is IgE already attached to mast cells in the bronchial mucosa or lung tissue, heightening allergic reactivity. The safest management includes avoiding all nuts, chickpeas, and sesame seeds, with an epinephrine auto-injector for emergency use. This case underscores the importance of monitoring for food allergies in transplant recipients, especially when the donor has a history of allergies.

Abstract Introduction Lung involvement is common in ANCA-associated vasculitis (AAV), presenting in various forms: necrotizing granulomatous inflammation, tracheobronchial inflammation, pulmonary capillaritis and diffuse alveolar hemorrhage (DAH), interstitial lung disease (ILD), and asthma.1 Radiologic manifestations are diverse, often complicating diagnosis based on imaging alone. Typical patterns include lung nodules, cavitation, pulmonary hemorrhage, and peripheral consolidation; however, cystic-like lesions are rarely seen.2 We present a case of AAV with bilateral cystic-like lung lesions. Case presentation A 54-year-old male with childhood asthma initially presented to an outside hospital due to progressive exertional dyspnea for 6 years and dysphonia for 1 year. Initial computer tomography (CT) revealed pulmonary cysts (Figure 1), and a CT-guided biopsy of a right upper lobe cystic lesion revealed intra-alveolar hemorrhage, though pathology was non-diagnostic. He received a diagnosis of congenital cystic lung disease and prescribed Trelegy and fluticasone inhalers. He was instructed to take oral prednisone with taper whenever dyspnea recurs for presumed asthma exacerbation. His exertional dyspnea indeed respond to prednisone but not to inhalers, and it would recur a few weeks later, and his dysphonia persisted. He came to our facility for second opinion given being told lung transplantation is the only ultimate treatment in his case. Additional history revealed chronic ear infection, bilateral rupturesd tympanic membrane, chronic rhinosinusitis, and recurrent nasolacrimal duct infection. He had been on oral prednisone for 4 weeks and remained on 10mg daily. Physical exam noted inspiratory wheezing. Further workups were notable for normal eospinophils, normal renal function, positive C-ANCA and PR3, normal pulmonary function test. High-resolution CT deomnstrated numerous areas of parenchymal lucency with circumferential surrounding groundglass opacity (GGO). At this point, the diagnosis of PR3-AAV is made. Patient is subsequently referred to rheumatology for further management. Discussion The diagnosis of AAV relies on clinical, serologic, radiologic, and pathologic data. The atypical presentation may delay diagnosis, especially in cases limited to ear, nose, and throat involvement and the radiological findings are of low specificity.3 While cystic lesions in AAV-related interstitial pneumonia are documented4, this case lacked a defined cyst wall. The central lucency may present partially destroyed lung parenchyma due to GPA-related necrosis (emphysema or pneumatocele) with surrounding areas of blood products deposited from prior peri-lesional hemorrhage. Clinicians should consider a broad differentials when encounter a cystic-like lesion with thorough history-taking being essential.

Abstract Background:Chronic obstructive pulmonary disease (COPD) is a major global health burden, with genome-wide association studies explaining only a fraction of its heritability. Recent advancements, including large-scale, multi-ancestry whole-genome sequencing and a focus on severe, early-onset cases, offer promising avenues to identify additional COPD-related variants. Furthermore, recent studies show that adjusting for polygenic risk scores (PRS) enhances statistical power to detect rare variants, advancing efforts to understand COPD's genetic basis. Methods:We performed a whole genome sequencing association study of severe, early-onset COPD (FEV1 < 50%, age ≤ 55), FEV1 and FEV1/FVC in eight population-based studies (ARIC, CARDIA, CFS, CHS, FHS, HCHS/SOL, JHS, MESA) and seven COPD-enriched studies (COPDGene, ECLIPSE, Boston Early-Onset COPD Study, LTRC, TCGS (Poland), ICGN, SPIROMICS). We performed single-variant analyses using regenie software, applying a genome-wide significance threshold of P < 5.0 x 10-8. For rare variant set (MAF <1%) analysis, we annotated variants with loss of function or deleterious missense using WGS Annotator and prediction tools LOFTEE, SIFT, and PolyPhen2, then analyzed these variants with SKATO-ACAT method in regenie, adjusting for covariates including sex, sequencing center, and principal components of ancestry. We additionally evaluated adjusting the analyses for PRSs for lung function traits. Results: We analyzed 53,028 participants including 34,751 European, 12,508 African, 1,088 Latino American, and 674 East Asian ancestry. Across cohorts, we analyzed 5,063 severe (FEV1 < 50%) COPD cases, among whom 1,103 were early-onset (age ≤ 55). Excluding variants within 1 Mbp of known variants, single-variant analysis identified 68 novel associations (MAC ≥ 30), with 8 common associations (MAF ≥ 1%) in 6 genomic regions. Our most significant novel association for severe early-onset COPD was rs74297390 (intergenic to POM121L12, MAF = 9.6%, Beta = 0.41, P = 3.5 x 10-8). For FEV1 and FEV1/FVC ratio, variants near DKK2 and AKR1E2 were significantly associated and were replicated in an independent dataset (Shrine et al., 2019). DKK2 is an antagonist in the Wnt pathway. In a preliminary variant set analysis, CHRM3-AS1 was associated with severe early-onset COPD, GPAT with FEV1, and COL8A1I and HMCN1 with FEV1/FVC. With PRS adjustment, ST6GALNAC5 was associated with FEV1/FVC. Conclusions: In a whole genome sequencing study of lung function and severe early-onset COPD, we identified new loci for COPD-related phenotypes, including two novel loci that replicated in an independent dataset. Ongoing work involves variant set-based analyses with additional covariate adjustments, as well as PRS adjustments for binary traits.

Abstract Rationale Amikacin liposome inhalation suspension (ALIS), as part of a combination antibacterial regimen, is currently the only FDA-approved treatment for patients with refractory Mycobacterium avium complex lung disease (rMACLD). MAC is the most common cause of nontuberculous mycobacterial lung disease (NTMLD). A real-world evidence study demonstrated significant decreases in hospitalizations from baseline over a 12-month follow-up period following ALIS initiation; however, study conclusions were limited by the absence of a comparator arm. In this study, we compared healthcare resource utilization in US patients who initiated ALIS versus non-ALIS antibiotic treatment for possible rMACLD. Methods This was a retrospective claims analysis using the Merative™ MarketScan® databases between 2016 and 2021. As no diagnosis codes for MACLD exist in claims, we defined 2 cohorts of patients receiving treatment for possible rMACLD as follows: 1) patients who initiated ≥28 days of continuous treatment with ALIS; and 2) patients who initiated ≥28 days of continuous treatment with ≥1 non-ALIS antibiotic used to treat rMACLD (i.e., parenteral aminoglycoside, clofazimine, fluoroquinolones, linezolid, tedizolid, or bedaquiline) ≥6 months after initiating treatment used for MACLD (i.e., with a combination of ≥2 antibiotics from a macrolide, ethambutol, and a rifamycin). The date of initiating treatment with ALIS or non-ALIS antibiotic for rMACLD was referred to as the index date. Other inclusion criteria: ≥1 diagnosis code for NTMLD anytime pre-index, ≥6 months continuous eligibility pre-index, and ≥12 months continuous eligibility post-index. Hospitalizations (all-cause, respiratory-related, NTMLD-related) and emergency room (ER) visits at 0-6 and 7-12 month post-index periods were compared with baseline (6 month pre-index period) per treatment arm. Results A total of 116 patients comprised the ALIS cohort and 63 patients the non-ALIS cohort. In the ALIS cohort, statistically significant reductions from baseline in all-cause, respiratory-related, and NTMLD-related hospitalizations were observed during the 7-12 month post-index period (Table). In contrast, the non-ALIS cohort did not show significant reductions from baseline in hospitalizations at follow-up. No statistically significant changes from baseline in ER visits were observed in either group. Conclusions Patients treated with ALIS for rMACLD had statistically significant reductions in hospitalizations from baseline during the 7-12 month post-index period, whereas no statistically significant improvements in hospitalizations were observed in the non-ALIS antibiotics group.