By Dr Matthew Callister, Consultant Respiratory Physician
A year after the preliminary publication of the NELSON screening trial results at the World Conference on Lung Cancer 2018 in Toronto, low-dose CT screening was again headline news in the Presidential Symposium at the WCLC 2019 in Barcelona. With the NELSON paper now submitted for publication, and many nations outside North America now actively planning or delivering lung cancer screening programmes, attention is focusing on aspects of implementation. Three presentations in the symposium presented data on who best to invite for screening, and the extent to which biomarkers can improve the efficiency of the screening process.
Preliminary data was presented from 4,985 participants in the International Lung Screen Trial (ILST), a collaboration between centres in Canada, Australia, Hong Kong and the UK (data from the UCL Lung Screen Uptake Trial was included in the analysis). The proportion of screening participants found to have lung cancer was compared for people selected for screening by the US Preventive Services Task Force criteria (the current criteria used in the US based on age and smoking history) versus a lung cancer risk prediction tool called the PLCOM2012 model. This uses 11 parameters to derive a lung cancer risk over the following 6 years. The presented data showed that, using a PLCO threshold that selected equivalent numbers of people for screening, the PLCO model detected 105 cancers compared to 85 using the USPTSF criteria (a 24% increase), suggesting it may be a more efficient method of selecting people for screening. The ILST is now entering its second round of screening; in addition, two large on-going studies in the UK are also comparing these criteria (the Summit study and the Yorkshire Lung Screening Trial) so more data will be forthcoming in due course. The only proviso to using tools such as PLCO is that they tend to select older people for screening compared to the USPSTF criteria, so that although more cancers are detected, the effect on life years gained may be somewhat attenuated.
By far the largest media interest was for the presentation of the first results of the Early Detection of Lung Cancer in Scotland (ECLS) trial, with headline articles in major UK news outlets. This study randomised 12,215 participants at high risk of lung cancer to either usual care or an intervention which comprised the EarlyCDT®-Lung Test followed by 6-monthly CT screening only if the test was positive. EarlyCDT®-Lung is a panel of seven autoantibodies which has been shown to have a sensitivity of 41% and specificity of 90% for lung cancer in previous analyses. The primary outcome measure was the rate of advanced lung cancer in the two groups, and a significant reduction was seen in the intervention group (HR 0.64, 95%CI 0.41-0.99). This led to headlines such as “New blood test that could dramatically reduce deaths from lung cancer” (Daily Telegraph) and “Ground-breaking blood test could reduce lung cancer deaths” (Sky News). However, to assess the utility of a blood test used in this way, one needs to be able to tease apart the benefit of the blood test itself over and above the already-proven benefit of LDCT screening. The stage-shift associated with CT screening is now well established; a similarly designed study whereby people at high risk of lung cancer are selected for screening by flipping a coin might demonstrate an overall stage shift compared to a control population. In order to show utility, the EarlyCDT®-Lung test would need to demonstrate enrichment of cancers within the test positive group, thus the sensitivity of the test for lung cancer becomes particularly important.
The study reported that of 56 patients diagnosed with cancer in the intervention group within 2 years of randomisation, 18 had a positive blood test (and were therefore presumably picked up by LDCT screening). By implication, the other 38 cancers diagnosed in test-negative patients were presumably detected following symptomatic presentation. This represents a sensitivity of 32%, but even this might be an over-estimate as sufficient time has not yet elapsed for all the non-screen detected cancers to present in the test-negative group. This compares to a sensitivity estimate of 44% for PLCOM2012 (with a threshold for screening eligibility of 1.51%) in a UK lung cancer population. Also of note is the overall number of lung cancers diagnosed in the intervention versus control groups. Screening works by bringing diagnoses of cancers forward in time, so one would always expect increased numbers of cancers in the screened versus control group early in any screening study, with catch-up of the control group later on. The fact therefore that fewer cancers were diagnosed in the ECLS intervention group compared to control group within 2 years of randomisation (56 versus 71) is unexpected. Data from a longer follow-up period and the published manuscript for ECLS are therefore awaited with interest to allow us to determine how and where this blood test might improve the screening process.
The third presentation related to screening was from the Italian BioMILD study. The MILD study has already reported its screening outcomes, demonstrating a 39% reduction in lung cancer mortality by prolonged LDCT screening beyond 5 years. The data presented in Barcelona related to the additional value of blood microRNA (miRNA) alongside LDCT in 4,119 participants. Unlike the ELCS trial, participants were scanned irrespective of the biomarker result, with the miRNA used to provide an additional screening outcome alongside the CT result. Those participants with negative LDCT and negative miRNA were recalled for their next CT at 3 years from baseline. Those with either a non-negative CT result (i.e. indeterminate or positive), a positive miRNA test, or both were recalled for annual or shorter repeat scan. Cumulative lung cancer incidence after 4 rounds of screening was 0.6% for the double negative group (negative CT and miRNA – termed 2neg), 3.8% where either miRNA was positive or LDCT non-negative (termed 1pos), and 20.1% where miRNA was positive and LDCT non-negative (termed 2pos).
The “1pos” group includes both those with positive miRNA/negative CT and those with negative miRNA/non-negative CT. Clearly separating these groups would be of great interest in determining the specific contribution of the miRNA test, over and above the information provided by the CT. However, one of the most interesting findings was the lack of harm in the 58% of participants falling in the “2neg” group left for 3 years. The stage distribution and interval cancer incidence were no worse in this group than in the higher-risk participants who were followed more closely. This contrasts with results of prior efforts to lengthen screening intervals beyond 2 years (such as in the NELSON trial), suggesting that biomarkers may help identify a low risk group where screening frequency can be safely lengthened. For those designing national screening programmes who are currently totting up the number of scans (and radiologists) required to deliver comprehensive population-based screening, this will be of particular interest.
The conference as a whole was characterised by huge optimism in terms of advances being made in lung cancer care. Rooms were packed with delegates hearing about newer combinations of oncological treatments, new data to guide surgical management, and other advances in screening to detect lung cancer at its earliest stage. Unfortunately though, sessions on tobacco control were much less popular. Of over 7,000 delegates attending the conference, less than 30 were in the room to hear the IASLC Lectureship Award for Tobacco Control given by Emily Stone from Sydney. Australia leads the world in its efforts to reduce smoking prevalence, and the rest of the world has much to learn from them in terms of the most effective and cost-effective way to reduce lung cancer mortality.
I am grateful to Hilary Robbins, Richard Booton, Emma O’Dowd, Sam Janes and David Baldwin for reviewing and editing this blog.