Clinical profile for future bronchiectasis on LCA
We identified one (100.0%) distinct clinical profile for Indigenous infants at-risk for future bronchiectasis (Table-4) and one without any risk (0.0%) (Profile-B). Infants in Profile-C who all had bronchiectasis also had the highest frequency of markers of severity for bronchiolitis i.e. supplemental oxygen requirement (100.0%), moderate accessory muscle use (84.7%), any co-morbidity (75.4%) and any bacteria (93.1%). Profile-A was the second most severe group with bronchiectasis in 35.4%, preterm birth in 90.7%, low birth weight in 89.2% and previous respiratory hospitalisation in 39.6%. However, many infants in this profile lacked known markers of severity (e.g. prolonged LOS, supplemental oxygen requirement etc).
\sout
DISCUSSION
In this first study using LCA in a cohort at-risk of bronchiectasis rather than asthma, we identified 5 distinct clinical profiles among the 164 Indigenous infants hospitalised with bronchiolitis. The most striking characteristics were Profile-A phenotype was being preterm (90.7%), low birth-weight (89.2%) and weight-for-length z-score <-1 (82.7%). Profile-B phenotype were characterised by all (100.0%) requiring oxygen supplementation (100%) yet absence of future chronic symptoms necessitating chest CT scan (0.0%) whilst all in Profile-C had future bronchiectasis (100.0%) and at the point of hospitalisation, most (93.0%) had NP bacteria. Profile-D was characterised by the absence of low birth-weight (100.0%) yet low weight for length z-score of <-2 (36.0%) and hRV infection (49.4%) and infants in Profile-E had low concurrent co-morbidity (33.9%).
Use of clinical phenotypes for various diseases have been appreciated increasingly over the last decade in a range of conditions. LCA has been generally used for this and to the best of our knowledge, there are only two such published studies3,4 involving infants with bronchiolitis. Both these novel studies3,4 were based in high-income settings where asthma is the outcome of interest. In the absence of any such studies in a setting where infants are at high risk of chronic suppurative lung disease30, we undertook this study to determine of LCA can identify phenotypes that may inform future interventions relevant to children at high-risk of bronchiectasis.
Uniquely, we identified two main clinical profiles with future bronchiectasis i.e. all in Profile-C had bronchiectasis and the second most common group with bronchiectasis was Profile-A (35.0%). Over a third of infants were included in these profiles which included many known risk factors for severe disease. It was not surprising that these profiles included the greatest number of infants with bacterial carriage and current co-morbidities. In Australia, Indigenous children particularly from remote communities often live in overcrowded homes30, are exposed to early and dense acquisition of respiratory bacteria (Streptococcus pneumoniae, Haemophilus influenzae and Moraxella catarrhalis )31 which are associated with poorer clinical outcomes (chronic wet cough18 and bronchiectasis30) in our setting. Whilst the above finding need confirmation in a large cohort from multi-centres, our novel findings highlight that LCA could be used for phenotyping children at-risk of future bronchiectasis and subsequently inform targeted interventions for these infants that could possibly prevent future bronchiectasis.
Unsurprising, our phenotypes are very different from that of Dumas and colleagues’3 for many reasons, of which the most important is the different target population. Further our study was substantially smaller, we excluded infants admitted to the intensive care unit, and did not include all the variables examined by Dumas and colleagues3. Nevertheless, we identified 5 distinct clinical profiles by use of LCA. Profile-B was the third largest group with markers of severity (e.g. marked accessory muscle use, oxygen requirement and prolonged LOS). This group is common to that of the USA cohort,3 whereby Dumas and colleagues3 reported a severe profile among infants with RSV, moderate to severe accessory muscle use and prolonged LOS (>3-days). The same result however was not replicated in the Finish study3 where these markers of severity (e.g. retractions and LOS) were distributed across two profiles. However, our Profile-B included oxygen requirement, a factor that was not included in Dumas and colleagues’3 LCA profile.
The limitations of the original studies whereby this cohort was obtained from, were previously published19-21. However, this study has further other important limitations. Firstly, we did not use BIC because our relatively small sample size restricted the use of BIC as cohorts of <400 that BIC underestimate classes32. We thus used aBIC, suitable for smaller cohort26,32. As aBIC still suggested that the sample is sufficiently large (n≥200)32, we thus included entropy as a second information criterion33 in class-models. Secondly, our data was cross-sectional and weakened by the lack of longer follow-up. Thirdly, our data were limited to Indigenous hospitalised for bronchiolitis with the lack of community-based children and the absence of recurrent wheezing and asthma, important factors in studies from high-income settings34.
In conclusion, our study is unique as, for the first time, we identified 5 clinical phenotypes in an at-risk population by using LCA. Two profiles were important for future bronchiectasis. Importantly, our data further highlights the heterogeneity of bronchiolitis phenotypes among infants. We now need confirmation of our novel findings in a large multi-centre cohort to determine such phenotypes in children at-risk of future bronchiectasis. Such work could subsequently inform targeted interventions for these infants that could possibly prevent future bronchiectasis, an increasingly recognised condition worldwide of which a significant proportion commences in childhood13.