EarlyCDT Lung for assessing risk of lung cancer in solid lung nodules

The external assessment group's (EAG) systematic review found only 5 study populations in which EarlyCDT Lung data on people with nodules was reported. Of the 5 studies, 3 were conference abstracts and only 2 were reported in full. The 2 complete publications had a high risk of bias in terms of patient selection, and the German study (Gonzalez et al. 2021) was also at high risk of bias in the flow and timing domain of the quality assessment. There were also serious concerns about the applicability of the results of these studies to NHS practice because of the position in the pathway where the test was used and how the test was used and interpreted. The EAG's meta-analysis estimated that EarlyCDT Lung has a sensitivity of 20.2% (95% confidence interval [CI] 10.5 to 35.5) and specificity of 92.2% (95% CI 86.2 to 95.8). The committee discussed that the EAG's estimate was uncertain with wide confidence intervals for the sensitivity. It noted that these results are different to the company's estimates, which were based on a single study that had 41.3% sensitivity (95% CI 35.0 to 47.6) and 90.6% specificity (95% CI 87.1 to 94.1; Healey et al. 2017). The EAG suggest that the inclusion of case-control data in Healey et al. (2017), where cases were people diagnosed with lung cancer and controls were healthy volunteers, could account for the difference in accuracy estimates. It suggested that poorer diagnostic accuracy among smaller nodules, which are less likely to be present in the confirmed cancer cases, could bias the results. The committee discussed if accuracy data from screening studies was generalisable to this decision question. The EAG commented that the malignancy risk is likely to differ between a screening population – that is people who may or may not have nodules – and a population of people who have nodules. Therefore, the screening data cannot be extrapolated reliably. The committee concluded that given the uncertainties around the accuracy estimates and the populations studied, further evidence of the accuracy of EarlyCDT Lung in people with nodules is needed (see section 4.1).

The committee understood that positive EarlyCDT Lung results are intended to be combined with a pre-test risk (using existing malignancy risk assessment models) to give a post-test risk. The EAG's report highlighted that the risk model (Healey et al. 2017) provided by the company to recalculate risk scores, may overestimate diagnostic accuracy as it was based on evidence from case-control studies (see section 3.2). The committee agreed that combining the tests was an appropriate way to estimate malignancy risk, but also noted that the accuracy data on these test combinations is very limited. The committee recommended that further validation of the model that combines the EarlyCDT Lung result with existing risk models is needed (see section 4.1).

The EAG's report found no evidence on how EarlyCDT Lung testing affects changes in clinical decision making for people with lung nodules. The EAG's simulation study indicated that EarlyCDT Lung is unlikely to significantly impact clinical management in low-risk nodules (below 10%). EarlyCDT Lung may improve clinical management in intermediate-risk nodules, particularly for those with a higher pre-test risk (above 48%). The data feeding into the simulations however was at high risk of bias, and the simulation had to use assumptions because of weak evidence. The committee discussed that data is needed to demonstrate if a positive EarlyCDT Lung test result changes decision making such as moving from CT surveillance to biopsy or from biopsy to immediate excision without biopsy (see section 4.1).

The EAG and committee discussed the risk and harms of false-positive test results (benign nodules incorrectly identified as malignant). The committee discussed that the safety of biopsy and surgical excision has improved in recent years, but it noted that there may still be adverse effects and anxiety for patients. It also discussed the harms of diagnoses in people with limited life expectancy or with multiple comorbidities, for whom treatment is not possible but the test could cause unnecessary anxiety. The committee discussed that data on the harms of biopsy and excision should be available through existing databases and recommended that it could be obtained as part of a large audit (see section 4.2).

In the EAG report no studies of EarlyCDT Lung in the target population reported health-related quality of life outcomes. One screening study (Early Detection of Cancer of the Lung Scotland trial) did report that there were no statistically significant differences in lung cancer worry, health anxiety, illness perceptions, lung cancer risk perception or intrusive thoughts, between people with and without lung nodules at 3 months and 6 months. The committee heard that the ongoing Artificial intelligence and Big Data for Early Lung Cancer Diagnosis (IDEAL) study is exploring the quality of life of people with lung nodules on CT surveillance. The committee concluded that EarlyCDT Lung test results could have an impact on anxiety, but that no evidence is available to support this.

No direct trial evidence was found on how EarlyCDT Lung impacts on long-term patient outcomes, such as lung cancer related mortality and morbidity, morbidity associated with other diagnostic tests or procedures, and overall and disease-free survival. The committee noted that this was an important factor to understand, but heard that a randomised control trial for assessing the impact of EarlyCDT Lung on patient outcomes would not be feasible because of the number of patients needed to power the study. The committee agreed that a linked-evidence modelling approach could be used to estimate the impact of EarlyCDT Lung on long-term patient outcomes. This method would link new diagnostic accuracy and decision impact data to existing clinical outcome data on lung nodules and lung cancer (see section 4.1).

The EAG's report highlighted variation in management practices and a lack of evidence around decisions in managing intermediate nodules. Under the British Thoracic Society's (BTS) guidelines for the investigation and management of pulmonary nodules (2015) people with nodules in the intermediate 10% to 70% risk group may be offered CT surveillance, biopsy or surgical excision. A clinical expert explained that in the BTS guidelines, the defined malignancy risk categories for initiating Brock and Herder risk assessment, and for guiding further testing and treatment, were mostly based on evidence graded at a level 3 (non-analytical studies; for example, case reports, case series). The committee recommended that an audit of existing data should be carried out to determine how these nodules are currently managed in NHS practice. It considered that the impact of EarlyCDT Lung and other tests would be difficult to ascertain without first understanding what happens in current clinical practice (see section 4.2).

The committee noted that the potential benefit of EarlyCDT Lung is that it could identify malignant lung nodules earlier than standard tests such as CT surveillance. This could lead to earlier treatment and improved patient outcomes. It also noted that there was no evidence that stage progression of malignant nodules happens in the timeframe of CT surveillance. The EAG commented that modelling studies of lung cancer screening strategies often evaluate the likelihood of pre-clinical stage progression over time, but cautioned that the generalisability of this evidence is unclear. The committee noted that it is important to understand if and how lung nodules progress during CT surveillance. This information could then be used in a linked-evidence modelling approach to understand if EarlyCDT Lung could result in meaningful earlier diagnosis of malignant lung nodules. The committee discussed that this data should be available through existing databases and recommended that it could be obtained as part of a large audit (see section 4.2).

The EAG report highlighted that better characterisation of the target population and nodule characteristics is needed for future assessments. It noted that the group of people with lung nodules is heterogeneous and different factors may impact on the speed of progression of malignant lung nodules. The committee agreed that this is important data to collect because it could be used in future modelling of the clinical and cost effectiveness of EarlyCDT Lung or other new tests in this area (see section 4.2).

The committee noted that a better understanding of the population with lung nodules and of the current diagnostic pathway is critical for supporting a linked-evidence model for EarlyCDT Lung and for other new technologies that would be used in the same pathway (see section 2.3). These issues include:

• How patient and nodule characteristics impact disease progression (see section 3.11).

• How intermediate-risk nodules are currently managed (see section 3.9).

• The clinical consequences of CT surveillance of lung nodules (see section 3.10).

• The prevalence, likelihood and impact of false-positive test results and overdiagnosis (see section 3.6).