Potentially serious incidental findings in the UK Biobank Imaging Study

Abstract

The increased use of imaging across research, clinical and commercial contexts has generated debate and calls for evidence on the benefits and harms of incidental findings (defined as those which are unrelated to the purpose of imaging) to inform policy and practice. Evidence on clearly non-serious incidental findings is of limited clinical usefulness; this thesis therefore focuses on potentially serious incidental findings (PSIFs), defined as those which may indicate the possibility of a condition which, if it was confirmed, would carry a real prospect of seriously threatening life span, or of having a substantial impact on major body functions or quality of life.

In 2014, the UK Biobank Imaging Study began performing brain, cardiac and body magnetic resonance imaging (MRI), dual-energy X-ray absorptiometry and carotid Doppler ultrasound, and aims to image 100,000 of its population-based participants. The imaging data can be combined with extensive sociodemographic, lifestyle, physical measures, biochemical, genetic and linked healthcare data, to generate a research resource which will facilitate studies into a wide range of diseases. Due to the scale of the UK Biobank Imaging Study, PSIFs are a particularly pertinent issue. UK Biobank therefore evaluates the impact of its protocol for handling PSIFs, the data from which form the basis of this thesis.

This thesis aims to provide empirical data on seven themes relating to PSIFs: their prevalence and nature; follow-up and final diagnoses; factors associated with PSIFs and with serious final diagnoses; participants’ understanding of consent to feedback of PSIFs; nonmedical impacts of feedback of PSIFs; opinions of receiving feedback of PSIFs; and the economic impact of feedback of PSIFs on hospital services.

Chapter 1 outlines the scale of the challenge of incidental findings, and summarises current literature and gaps in our knowledge relating to each of the seven themes on PSIFs. Chapter 2 reviews systematically and meta-analyses published studies of brain and body MRI of apparently asymptomatic adults. Chapter 3 introduces the UK Biobank, the UK Biobank Imaging Study, and the rationale behind and protocol used to handle PSIFs in 100,000 largely asymptomatic participants: radiographer flagging of concerning images for a radiologist to review. Chapter 4 presents a study comparing two protocols to handle PSIFs in the first 1,000 imaged UK Biobank participants: radiographer flagging versus systematic radiologist review of all images. Chapter 5 investigates the factors associated with PSIFs and with serious final diagnoses. Chapter 6 examines the economic impact of feedback of PSIFs on hospital services, using linked routinely collected healthcare data. In the systematic review, pooled prevalences of PSIFs on brain, thorax, abdominal and brain and body MRI were: 1.4–1.7%; 1.3–3.0%; 1.9–4.5%; and 3.9–12.8% respectively, the upper estimates reflecting the inclusion of indeterminate findings. There was substantial heterogeneity, but few informative data on potential sources of this. Around half of PSIFs were suspected malignancies.

Based on the first 7,334 participants in the UK Biobank Imaging Study (283 of whom had PSIFs), the PSIFs protocol had the largest influence on the prevalence of PSIFs and serious final diagnoses of any of the investigated factors: systematic radiologist review resulted in around 13 times more PSIFs and around four times more serious final diagnoses compared to radiographer flagging. A lower proportion of PSIFs detected by radiologists were finally diagnosed as serious compared to radiographer flagging (12% and 32% [Chapter 4 and 5]).

Feedback of PSIFs resulted in substantial impacts in terms of: clinical assessments (all participants visited their general practitioner, and 90% underwent some form of other clinical assessment, mostly imaging or referral to a specialist [Chapter 4]); non-medical impacts on participants (including on emotional wellbeing, insurance and finances and work and activities in 17%, 9% and 6% respectively [Chapter 4]); and hospital service use and cost (81% of cases with PSIFs generated some hospital use and costs, which had increased compared to controls, and to cases’ hospital use and costs during the year before feedback of a PSIF [Chapter 6]). Importantly, as around 80% of PSIFs turned out not to be serious (Chapters 2, 4 and 5), many of these impacts may be unnecessary.

Despite these negative impacts, the vast majority of participants were glad to have received feedback of a PSIF and to have taken part in the imaging study (98% and 99% respectively), although almost a quarter changed their minds over time about whether or not feedback should always be given. Around a quarter of participants incorrectly thought they could choose to receive feedback and UK Biobank has improved its consent materials accordingly (Chapter 4).

Feedback of PSIFs impacts on participants and publicly-funded health services (and in turn patients in need); most PSIFs turn out not to be serious and many of these impacts may be unnecessary. Researchers can substantially influence these impacts via IFs policies, which must be designed to minimise unnecessary harms, and be clearly explained to participants to facilitate informed consent. These, and other implications of this thesis are further described in Chapter 7, which also discusses the results in the context of the broader literature, outlines the strengths and limitations of this thesis, and suggests directions for future work.