Repository logo
 

Understanding and predicting acute pulmonary exacerbations in adults with cystic fibrosis


Type

Thesis

Change log

Authors

Ukor, Emem-Fong 

Abstract

ABSTRACT

Understanding and predicting acute pulmonary exacerbations in adults with cystic fibrosis

Dr Emem-Fong Ukor

Cystic fibrosis (CF) is the most common, inherited, life-limiting, multi-system disorder among the Caucasian population. Disease arises from mutations in the cystic fibrosis trans-membrane conductance regulator (CFTR) gene. In the lung, defective protein leads to the accumulation of thick viscid mucus, depletion and acidification of the airway surface liquid and impairment of the normal muco-ciliary clearance mechanisms, providing a permissive environment for chronic infection and progressive airway damage. Pseudomonas aeruginosa is a versatile, Gram-negative opportunistic human pathogen with a predilection for establishing chronic infection in the CF lung. Its extraordinary capacity to cause infections is due to its vast repertoire of secreted and cell-associated virulence determinants, which are subject to a complex regulatory network of intracellular and intercellular signals. Acute pulmonary exacerbations (APE), are the main cause of morbidity and mortality in CF. Despite their clinical significance, the mechanisms that trigger these events are poorly understood. In this dissertation, I investigate whether home monitoring for changes in patient physiology and symptoms was feasible and could permit advanced detection of an APE. I additionally concentrated on whether temporal fluctuations in the behaviour and structure of established P. aeruginosa populations within the CF lung may trigger APEs, and whether such changes could function as a bacterial biomarker(s) and be correlated with home monitored data to facilitate APE diagnosis and prompt initiation of treatment.

First, I conducted a single-centre, pilot study (TeleCF) of 15 adults with CF in order to determine whether daily home monitoring of a single sputum bacterial biomarker (P. aeruginosa exotoxin A [PEA]) along with several clinical parameters might provide advanced warning of an APE. Home monitoring was well tolerated and provided high resolution data on physiological and biomarker changes preceding, during and following antibiotic therapy for an APE. On its own, PEA did not prove an effective biomarker for early detection of APEs in adults with CF, but the study did provide proof of concept for the application of home-monitoring and sputum profiling for bacterial biomarkers to inform further work.

Next, I collected longitudinally sampled sets of 95 isolates per sputum sample from 9 adults with CF before, during and after antibiotic treatment for an APE. Isolate populations were analysed for a series of phenotypic traits associated with P. aeruginosa virulence to determine whether changes in phenotype composition were related to exacerbation onset. I also investigated for differences in phenotype composition between isolate populations of the P. aeruginosa Liverpool epidemic strain (LES), Manchester epidemic strain (MES) and local non-epidemic strains. I found strong evidence for the uncoupling of the traditional quorum sensing (QS) regulatory hierarchy in CF isolates, with the rhl subsystem playing a more dominant role in virulence expression in certain strain types. Importantly, no link was found between APEs and the emergence of a particular sub-population of morphotypic or phenotypic variants.

Finally, I conducted a multicentre UK-based study (SMARTCARE) of 147 adults with CF to assess the acceptability and feasibility of daily monitoring of symptoms and physiology using novel sensor technology and mobile phones. Linked-anonymised data were analysed using machine learning (ML) methods to define the profile of APEs and predict their onset. Survey patient feedback confirmed that home monitoring was easy to do and helped patients track their health over time. Unsupervised machine learning analysis uncovered the typical signal profile of an APE and revealed three distinct classes of APE. We developed an ML predictive classifier that can detect an impending APE on average 11 days earlier than current clinical practice.

This work has contributed greater insights into the day-to-day variation in symptoms and physiology prior to, during and following periods of APE in adults with CF. It has confirmed the important role for home monitoring in CF care delivery and highlighted the power of machine learning methods when applied to high frequency data to advance our understanding of APEs in adults with CF.

Description

Date

2021-04-01

Advisors

Floto, Andres

Keywords

pulmonary exacerbations, cystic fibrosis, biomarkers, machine learning, Pseudomonas aeruginosa, telemonitoring, phenotypic analysis, prediction, home monitoring, virulence factors, quorum sensing, bluetooth technology

Qualification

Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge