Participant characteristics
A total of 337 bronchoscopy procedures were performed solely for research purposes as part of five NIH-funded studies from 1 January 2019 to 15 March 2022 and were included in the analysis. The pertinent characteristics of the study population are described in Table 1.
Table 1
Characteristics of the study population stratified by clinical group.
| Community Controls (n = 150) | Household Contacts (n = 57) | TB Pre-Treatment (n = 24) | TB Early Treatment (n = 54) | TB End of Treatment (n = 22) | PLHIV (n = 29) | Pooled (n = 337) |
Age in years | 29 (16.8) | 34 (17.8) | 32 (15.3) | 33 (21.5) | 31 (17.3) | 44 (13.5)* | 32 (20) |
Female | 89 (59.3) | 36 (66.7) | 7 (35.0) | 14 (25.9) | 6 (27.3) | 25 (86.0)* | 178 (54.1) |
Current tobacco smoking | 101 (67.3) | 31 (57.4) | 18 (90.0) | 40 (74.0) | 15 (68.2) | 7 (24.1)* | 212 (64.4) |
Recreational substance use | 21 (14.0) | 14 (25.9) | 6 (30.0) | 19 (35.2) | 7 (31.8) | 2 (6.9) | 70 (21.3) |
SARS-CoV-2 positive+ | 17 (11.3) | 14 (24.6) | 6 (25.0) | 0 | 0 | 6 (20.7) | 43 (15.9) |
Other comorbidity$ | 24 (16) | 9 (16.7) | 1 (5.0) | 8 (14.8) | 1 (4.5) | 4 (13.8) | 47 (13.9) |
IGRA status# positive negative | 101 (67.3) 41 (27.3) | 48 (88.9) 6 (11.1) | ND | ND | ND | 21 (72.4) 8 (27.6) | 184 (55.9) 57 (17.3) |
Age is expressed as median (interquartile range); all other variables are expressed as number followed by the percentage (in parentheses) that this number represents of valid samples in the specific study group for the specific variable.
*Two-sided Cochran-Armitage tests found no statistically significant differences between the first five clinical groups. On ANOVA, age was significantly different between PLHIV and the other clinical groups (P < 0.001, F-statistic = 7.842). Fisher’s exact test found that the PLHIV group had significantly more females (P < 0.001, 99.5% C.I. for odds ratio 1.77–107.9) and fewer current smokers (P < 0.001, 99.5% C.I. for odds ratio 0.03–0.5) than the other clinical groups.
+One study only performed SARS-CoV-2 RT-PCR, not serology, and thus participants were unable to be classified. The unclassified participants were not included in the analysis for this group.
$Excluding TB, HIV, and previous SARS-CoV-2 infection.
#Several participants’ IGRA results were ‘indeterminate’ or unknown; IGRA was not performed for participants with active TB.
IGRA, interferon gamma release assay; ND, not determined; PLHIV, people living with HIV; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2;TB, tuberculosis .
The most common comorbidity was hypertension (eight in Community Controls, six in Household Contacts, one in the TB groups and four in PLHIV). Four participants had asthma, four had previous chest trauma, and three had allergic rhinitis. The remaining comorbidities were a wide range of minor ailments. Approximately one fifth of the population used recreational substances, of whom 59 (85.5%) smoked cannabis (only five of these did not also smoke tobacco); four (5.8%) used both cannabis and another substance such as methamphetamine or methaqualone (all current tobacco smokers); and six (8.7%) used methamphetamine or methaqualone alone or in combination (all current tobacco smokers).
Of the 78 participants with active TB (the TB Pre-Treatment and TB Early Treatment groups), the mean Body Mass Index (BMI) was 19.5 (14.8–29.0), with 39 (50%) of participants classifying as ‘underweight’ and only four (5.1%) as ‘overweight’. On study entry, 39 (50.0%) participants were sputum smear positive, with scores as follows: five (6.4%) ‘scanty’ (less than 10 AFB in all fields); 17 (21.8%) ‘1+’ (less than 1 AFB per field); 13 (16.7%) ‘2+’ (1 to 10 AFB per field); and four (5.1%) ‘3+’ (more than 10 AFB per field). Of the 76 who had sputum MGIT culture, 71 (93.4%) were culture-positive, with a mean time to positivity of 10.6 days. Of the 70 that had sputum Xpert Ultra results, 66 (84.6%) were positive, with the following ranking: 17 (24.3%) ‘high’; nine (12.9%) ‘medium’; 13 (18.6%) ‘low’; six (8.6%) ‘very low’; three (4.3%) ‘trace’; 18 unknown. All participants were symptomatic of TB, but well enough to undergo bronchoscopy. All chest radiographs in the active TB groups were highly suggestive of active TB, except for one participant with a normal chest radiograph. In the TB End of Treatment group, 14 (63.6%) had abnormal radiographs not in keeping with active TB; four (18.2%) had radiographs suggestive of active TB; and only four (18.2%) had normal radiographs. Of the Household Contacts group, 20 (35.1%) had abnormal chest imaging (including both chest radiograph and PET-CT) which was not suggestive of TB, and 11 (19.3%) had chest imaging which was suggestive of TB (either related to previous TB episodes, or implying possible subclinical TB, as participants were asymptomatic and tested negative for TB on sputum). Only nine (6.0%) Community Controls had abnormal imaging (radiographs and PET-CTs), none were suggestive of TB. There were 28 PLHIV included, one of whom had two procedures. Using a composite of both chest imaging modalities, 12 (42.9%) of the PLHIV group were normal, three (10.7%) had results consistent with subclinical TB, 14 (50%) were abnormal but not in keeping with TB. All PLHIV were on antiretroviral therapy (ART) at the time of bronchoscopy, for a median of 10 years duration (range 0.5–17 y). Of the 18 participants who had CD4 counts within six months before the bronchoscopy, the median was 602 cells/µL. Of the 16 participants with HIV viral loads within six months, the median was < 20 copies per millilitre.
SARS-CoV-2 positive subgroup
The distribution of SARS-CoV-2 participants across clinical groups is shown in Table 1. Of the 42 participants (43 procedures), 31 (72.1%) were females, and the mean age was 36 years. There were similar proportions of current (20; 46.5%) and never (18;41.9%) smokers. The majority (37; 86.0%) did not use recreational substances, four (9.3%) used cannabis, and two (4.7%) used methamphetamines. Fourteen (32.6%) had a comorbidity, of which the most common was hypertension. Using a composite of both chest imaging modalities, 15 (34.9%) were normal, six (13.9%) were classified as in keeping with active TB, six (13.9%) had subclinical TB, and 16 (37.2%) were abnormal but not in keeping with active TB. Forty participants (95.2%) tested either anti-N antibody positive, anti-S antibody positive, or both. Four (9.5%) tested RT-qPCR positive in the three weeks prior to the bronchoscopy, however none of them had symptoms of coronavirus disease-2019 (Covid19) at the time of testing. Only 16 participants (38.1%) reported symptoms of Covid19 at any time (mean time between symptoms and bronchoscopy was 6.4 months; in 18 symptom status was unknown); two of these (4.8% of the group) required hospital admission; none of them had active TB.
BAL yields
Overall, the median BAL volume collected from the procedures was 130 ml (IQR 49 ml), which represents 54.2% of the instilled volume. The median number of BAL cells collected from the procedures was 30.2 x 106 (IQR 38.7 x 106). The median BAL cell concentration (cells/ml) was 0.24 x 106 (IQR 0.34 x 106). The median cell viability (proportion of live cells in the total cell yield) was high, at 96.8% (IQR 2.8%). The analysis showed no significant variability in BAL cell yield over time overall, or when stratified by parent study. There was no significant difference in BAL cell yields between operators performing the bronchoscopy, and no significant variability in their performance over time (Supplementary Fig. S1). We investigated the effect of the BAL volume on the total BAL cell yield and found that when the whole sample was considered there was no significant relationship between the two outcomes. However, when we performed group-specific testing, we found linear relationships in current smokers from the Household Contacts and Community Controls (P = 0.04) and in the TB treatment groups (p < 0.0001 for TB Early Treatment and P = 0.044 for TB Pre-Treatment. This suggests that the true relationship was confounded in sample population analysis, as shown in Fig. 2. There were 333 data points available for the pellet colour analysis: 215 (64.6%) pellets were black, 54 (16.2%) were clear, and 65 (19.5%) were bloody.
Factors affecting BAL yields
In a simple linear model assessing the impact of the available clinical variables on BAL volume yield, the factor with the greatest effect was the anatomical lung lobe chosen, with the RML yielding significantly higher volumes compared to all other lobes except the lingula (with RML as reference, RUL − 35.4 ml, P < 0.0001; LUL − 22.5 ml, P = 0.001; LUL − 30.8 ml, P = 0.022; RLL − 25.6 ml, P = 0.042; lingula − 12.0 ml, P = 0.198). Increasing age was associated with a lower volume yield (P < 0.0001), as was male gender (P = 0.0447) though the latter became non-significant after correction for multiple testing. When participant age was divided into three categories, pairwise comparisons (ANOVA on the medians) showed a relationship between these categories and BAL volume, cell yield and concentration, which remained significant after correction for multiple testing (Fig. 3).
Three variables were highly associated with total BAL cell yields: current smoking (26x106 cells more than never smokers, SE 7.5x106, P < 0.0001); male gender (16x106 cells more than female gender, SE 5.9x106, P = 0.007); and pellet colour (bloody pellet 19.6x106 fewer cells than black pellets, SE 7.3x106, P = 0.0119; clear pellet 20.1x106 fewer cells than black pellets, SE 9.0x106, P = 0.027). The findings for BAL cell concentration were similar. The effect of gender did not stand up to correction for multiple testing, and further interrogation found that the three variables smoking, gender and pellet colour, were highly correlated, explained by the fact that significantly more males were smokers, and smoking is the commonest cause of a black BAL pellet (Fig. 4).
To further investigate the effects of smoking on BAL characteristics, pairwise comparisons (ANOVA on the medians) were performed, showing clearly that although current smokers give the lowest BAL volume yields, they have the highest overall cell yields (Fig. 5). These effects were seen for the black BAL pellet category as well.
Pairwise comparisons (robust 2-way ANOVA on the medians) revealed some interesting trends in BAL volume and cell yields, from participants’ TB treatment status (Fig. 6).
The TB Early Treatment group and the TB End of Treatment group yielded lower BAL volumes than the Community Controls group [using the Community Controls as a reference, TB Early Treatment: -20.3 ml, 95% C.I. (-36.5, -4.06), P = 0.006; TB End of Treatment: -28.4 ml, 95% C.I. (-47.5, -9.28), P = 0.002], but the cell yields and concentrations in the TB Pre-Treatment group were higher than in the Community Controls (total BAL cells: 41.9x106 cells more than in Community Controls, 95%C.I. 18.6–68.8x106, P < 0.0001; BAL cell concentration: 0.44x106 cells/ml more than in Community Controls, 95%C.I. 0.25–0.66x106, P < 0.001). The TB Pre-Treatment group had a significantly higher cell yield (23.7x106 cells more than the TB End of Treatment group, 95%C.I. 7.45–40x106, P = 0.002) and concentration (0.2x106 cells/ml more than the TB End of Treatment group, 95%C.I. 0.01–0.4x106, P = 0.035) than the End of Treatment group, suggesting the TB treatment duration affected the cellular content of the BAL. The trend was confirmed on direct comparison of paired Early Treatment and End of Treatment samples (P = 0.025), and remained significant when the results were adjusted for smoking status. The apparent effect of HIV on lower BAL cell yields seen in Fig. 6 was confounded by most of these participants being never-smokers.
Variables which had no effect on the outcomes in the linear models included SARS-CoV-2 status, imaging findings (using a composite of chest radiograph and PET-CT when performed, categorised as ‘normal’, ‘abnormal, in keeping with active TB’, or ‘abnormal but not active TB’), any airway abnormalities observed during the procedure, IGRA result, recreational substance use, and the presence of other comorbidities. Interestingly, ANOVA on the medians revealed the following interactions: participants who had airways which collapsed on minimal suction during the procedure (suggesting underlying airways disease) had a lower volume yield than those with no observable airways abnormalities (P < 0.001); the presence of abnormal secretions (suggesting chronic lung disease or current infection) also predisposed to lower volume yield (P = 0.007); and both abnormal imaging categories (‘abnormal but not active TB’, and ‘abnormal, in keeping with active TB’) were associated with a lower BAL volume yield than those with normal imaging results (P = 0.004).
Factors affecting BAL pellet colour
ANOVA testing showed that the highest cell counts are found in black pellets, followed by bloody pellets, then clear pellets (P < 0.001). While red blood cells may obscure black particulate in pellets that would otherwise be labelled black (and thus explain the higher cell counts in bloody pellets), if the bloody pellet’s higher cell count represents contamination with blood immune cells, this has potentially negative implications for the post-thaw BAL cell viability and the interpretation of any BAL immunoassays. Logistic regression analysis found that being in the PLHIV group was associated with a bloody pellet (P = 0.013). As the only two parent studies to include PLHIV also had brushings during their procedures, we investigated whether the order in which the procedures were performed had any effect on the pellet colour and found that performing the brushing after the BAL (in the same lung lobe) was associated with a significantly lower rate of bloody BAL pellets compared to performing the brushing before the BAL (P = 0.001). As the reduced visibility and enhanced cough stimulus after the BAL may compromise participant safety when performing a brushing, our group opted to try a third technique in sampling the ipsilateral lower lobe before the BAL, but this was not successful in reducing the proportion of bloody pellets (Fig. 7).
BAL cell differential count
It has been shown that smaller instilled BAL volumes sample only the distal airways whereas larger volumes sample the alveolar compartment, resulting is differences in measured protein and cellular contents of BAL fluid.3 We therefore investigated whether the BAL volume yield affected cell differential counts, but found that there was no relationship between the two. On pairwise ANOVA on the medians, several significant differences in the differential cell counts between groups were found: 1) older age categories were associated with higher neutrophil proportions (P = 0.003); 2) the TB Early Treatment group had the highest proportions of both neutrophils and eosinophils of the clinical groups (P = 0.03 and P = 0.045, respectively); 3) clear BAL pellets had the highest proportions of lymphocytes (P = 0.002), black pellets had the highest proportions of neutrophils (P = 0.01), and bloody pellets did not have any significantly enriched cell populations; 4) never smokers had the highest proportions of lymphocytes (P = 0.045); ANOVA on means found that current smokers had significantly higher proportions of neutrophils and eosinophils, but this interaction was not significant on comparison of medians; and 5) SARS-CoV-2 positive participants had higher proportions of BAL lymphocytes than SARS-CoV-2 negative participants (P = 0.023). The median BAL differential cell counts in each clinical group are presented in Supplementary Table S2.
Adverse events
The mean procedure duration (counted from intubation with the bronchoscope to extubation) was 11.5 minutes (range 4–22), with a mode of 10 minutes. Procedures in which a brushing was done with the BAL were significantly longer than those without (P < 0.01); the two studies including brushings had more procedures lasting > 12 minutes (P < 0.05), > 14 minutes (P < 0.05) and > 15 minutes (P < 0.01).
There were no severe adverse events in this study. Minor adverse events were recorded for 118 (33.1%) participants. Supplementary Figure S2 shows the minor adverse events, stratified by clinical group. Minor bronchial mucosal bleeding was significantly more common in the PLHIV group (P < 0.001), all of whom also had brushings in the same procedure. Overall, 69 (23.5%) of 293 participants with available data reported still experiencing symptoms at 72 h after bronchoscopy (Supplementary Figure S3). ‘Throat discomfort’ and post procedure cough were significantly different between groups, being less frequently reported in the TB Pre-Treatment group (P = 0.0264), TB Early- and End of Treatment groups (P = 0.044) respectively. PLHIV were more likely to experience dizziness than other groups (P = 0.010).
Of those who experienced any symptom, eight (2.7%) sought medical attention for the problem. One (0.3%) study participant was admitted overnight to hospital, three (1.0%) received antibiotics, and six (2.0%) received other medications (including anti-inflammatories, antihistamines, steroids or analgesics). At the day 14 assessment, no participants reported any additional problems. Importantly, at the 72 h assessment, 234 (96.7%) of the 242 participants surveyed stated they would be willing to have a repeat procedure; 236 (97.5%) reported that their overall experience of the bronchoscopy was tolerable; and 239 (98.8%) participants reported that the bronchoscopy procedure was better than they expected. Results were similar at the day 14 assessment.