Derivation and validation of a novel model incorporating PET-CT for predicting malignancy in screen detected lung nodules - The PRECISE Study

Lung nodules are like freckles on the lung, and can be cancerous or not. They are increasingly found on CT scans. In the US there are approximately 1.5million Americans detected with lung nodules each year. Screening for lung cancer in high risk individuals using low dose CT is currently standard practice in the US and these scans find lung nodules in approximately 1 in 5 participants. This year, NHS England has announced a £70million project to pilot lung cancer screening in 7 areas in England which is estimated will detect new nodules in approximately 12000 patients.

Working out whether a nodule is a lung cancer (malignant) or not (benign) is therefore an important process. People with cancerous nodules should be offered prompt treatment (usually surgery) while those with benign nodules require no further assessment and can be reassured.

However, determining whether a nodule is benign or malignant is challenging and relies on estimating the chance that a nodule is malignant. Currently, this is done by utilising a risk calculator that includes a PET-CT scan. A PET-CT scan involves an injection of radioactive sugar and then observing whether the nodule (or anything else) takes up the sugar. The degree to which the nodule takes up the sugar correlates with the risk of malignancy. However, the risk calculator currently in use was developed in 2005 by studying only 106 patients. It has not been tested properly in patients undergoing screening for lung cancer and the scan technology has changed considerably since 2005. Worryingly, the current risk calculator underestimates the risk of cancer in lung nodules and a more accurate calculator is therefore urgently needed.

In this proposal we aim to develop a new risk calculator that will accurately predict the chance of a nodule being cancerous. The first step in this research will involve combining the data from PET-CT scans carried out in several trials of CT screening from lung cancer to develop a new risk calculator. Unidentifiable data on participants having PET-CT scans in these trials will be securely transferred to a new database at UCL. Over 10 factors will be examined such as family history, smoking, size of the nodule and appearance on PET-CT scan to see whether they predict the likelihood the nodule is malignant. Using these data and statistical methods a new risk calculator will be developed.

Once the new risk calculator has been developed it must be tested to see how it performs in a different group of patients before it can be used in clinical practice. Testing the new calculator will be carried out in the SUMMIT trial. This is a major trial which involves carrying out CT scans on 25000 individuals in London. It is estimated that this will result in approximately 5000 participants with nodules of which 750 will be cancerous. We will examine whether the new risk calculator can accurately predict whether the nodules found in the SUMMIT trial are cancerous or not. The new risk calculator will be compared to the existing calculator from 2005 as well as other calculators that do not include PET-CT. If the risk calculator is accurate then this may mean that patients will require fewer investigations at less cost, start treatments earlier and have less anxiety. It will be made available as an online tool to provide a likelihood that the nodule is cancerous, along with a recommendation for the next steps which can be be discussed by the individual and clinician.

The new risk calculator will also be tested in other scenarios. In particular it will be tested on nodules that are detected in routine practice (not in a screening programme) and also in patients who have several nodules found on CT scan. The impact of the risk calculator on patient pathways will be determined and may improve the management of lung nodules by reducing patient anxiety as well as the number and cost of future investigations.

Problem to be addressed: Pulmonary nodules are a common finding on CT scans and their frequency is increasing due to the implementation of lung cancer screening. In the US there are 1.5million people detected with lung nodules each year. Current nodule management estimates the probability that the nodule is malignant using risk prediction models. However, the existing Herder model (which includes PET) underestimates lung cancer risk and a more accurate model for predicting risk in pulmonary nodules is urgently required. Despite PET-CT being an expensive and limited resource, its role in nodule management has not been fully evaluated.

Research questions
1. The sensitivity, specificity, accuracy and cost-effectiveness of PET-CT in screen detected nodules (solid and part solid) using varying cut-offs of SUV threshold
2. The efficacy and cost-effectiveness of PET-CT prior to treatment of screen-detected nodules
3. The clinical utility and cost-effectiveness of a novel nodule risk model including PET-CT compared to current nodule risk predictors
4. The additional clinical utility and cost-effectiveness of PET-CT in patients with multiple pulmonary nodules

Methods:
An individual patient data (IPD) meta-analysis will be conducted of PET-CT in completed lung screening trials in the US, UK and Europe. This dataset of >8000 patients with nodules will be powered to generate a novel multivariable logistic-regression model to estimate the risk of lung cancer in screen-detected lung nodules, including predictors such as smoking exposure and nodule characteristics. The model derived from the screening trials will then be validated in the SUMMIT trial which aims to recruit 25000 individuals at high risk of lung cancer in London undergoing low dose CT chest by 2021. The cost-effectiveness of using the novel algorithm in a screened population will then be modelled. The model will also be tested in nodules detected in routine practice and in patients with multiple nodules.

The PRECISE research project will have impact on the following:

(i) Policy makers: results from the PRECISE model will be assessed by NICE and other relevant guidelines (British Thoracic Society in the UK and American College of Chest Physicians in the US). The model is expected to be superior to current models for predicting risk of malignancy and therefore is expected to have a short pathway to impact on national and international guidelines. A superior model is likely to be cost-effective and by reducing the number of scans required for follow-up will represent significant cost-savings to the healthcare system in the UK and globally.

(ii) The proposal is anticipated to have a significant impact on patients with lung nodules. This represents a growing number of people due to increased imaging in the UK and worldwide and the initiation of lung cancer screening in the US and many other countries. In the UK, lung cancer screening is being piloted by NHS England in 10 centres and will result in approximately 12000 people with screen detected nodules. It is expected that the results of this project will directly impact their care pathways. More accurate risk prediction of malignancy in nodules may improve patient understanding of nodules and reduce anxiety.

(iii) Clinical care will be directly impacted with potential for healthcare savings and reduction in patient anxiety. A superior risk model for lung nodules may reduce further scans with a reduction in radiation exposure and alleviate anxiety about future radiation induced cancers.

(iv) The new model will represent a significant advancement in scientific knowledge and form the new benchmark for future studies on nodule management and in particular the integration of serum and exhaled biomarkers.

(v) The work will emphasise the UK as a leader in lung cancer screening optimisation. Lung cancer remains the largest cause of cancer death worldwide and while screening has been shown to reduce mortality significantly, concerns regarding nodule management and downstream costs of managing nodules have prevented widespread implementation. The development of the PRECISE risk model has the potential to improve nodule management in a cost-effective way and therefore will encourage many healthcare systems to implement screening which has the potential to reduce lung cancer mortality by at least 20%.

(vi) The proposal will impact Dr Navani (PI) considerably. It will allow him to develop his research and leadership skills and give him the platform to apply for future funding at the end of this project with the aim of becoming a leader in clinical academic research.