Thread by @Garth_Nicholas1

21 November: Screening part 2

This year for Lung Cancer Awareness Month #LCAM I’m going to summarize 30 important lung cancer trials over 30 days. These posts are directed at non-medical professionals, with descriptions of the results and of what makes a good trial. #lcsm 1/17

Yesterday we went over the major results of the NLST. We discussed screening in general and the concept of overdiagnosis in particular. Today we’ll look at the Dutch-Belgian NELSON study, the next largest randomized study in this field. 2/17

NELSON enrolled 13 195 people between 2000 and 2004. They were randomly assigned to no screening, or to CT scans at baseline, 1 year, 3 years, and 5.5 years later. The trial was powered to detect a 25% reduction in lung cancer mortality over the 10 years from enrollment. 3/17

To be eligible for the trial, participants were:

1. Male
2. Smokers of >15 cigarettes/d for >25 years, or >10 cigs/d for >30 years
3. If quit, quit date <10 years ago

The requirement for male gender was criticized, and 2594 women were added later as a subgroup.
4/17

Follow up relied on the national health databases of the two countries, which contain information about new cancer diagnoses and dates of death. An expert committee met to adjudicate cause of death (recall death from lung cancer is the primary endpoint). 5/17

The trial also pioneered a novel approach of estimating the volume of lung nodules, and using the rate of change in volume over time to determine who needed additional workup. This substantially decreased the number of scans performed compared to NLST. 6/17

The primary outcome was met: The ratio of cancer deaths in the screened to non-screened arm was 0.76 (95% CI 0-61-0.94 p=0.01). This reflects 206 lung cancer deaths in the control arm, and 156 in the screened arm. Results in the women’s substudy were as favourable or moreso.

There was no observed improvement in overall survival, though, with the ratio of all-cause mortality in the two arms 1.01 (95% CI 0.92-1.11). The trial wasn’t powered to answer this question, but failure to demonstrate even a trend towards OS improvement must be a disappointment.

The authors are refreshingly candid about difficulties in establishing whether any particular death was due to lung cancer, with the members of the review committee being concordant 86.1% of the time. Occasional discordance with the official death certificates was also noted.

There are those who criticize these studies for their primary end point, for the accuracy of death attribution, for the x-rays in the control arm of NLST, for how long it took to publish NELSON. I think these criticisms miss the mark, and do not negate the observed benefits.10/17

There are also those who feel that the survival gains in the restricted populations enrolled in these studies could be expanded by making lung cancer screening available to broader populations. I think this also is probably not correct, at least not if done indiscriminately.11/17

I think that how you feel about screening depends a lot on what you think of the harms of screening, particularly overdiagnosis. Many people feel there is no harm in screening: “what’s the harm in a CT scan that could save your life?” 12/17

I hope I've convinced you that the harms of overdiagnosis are real. There are potential harms to the work up of benign lesions, and to radiation exposure itself. These harms may be small, but they show that screening in a population where lung cancer is rare may be net harmful.

Overdiagnosis should make us careful of the trial endpoints we accept. Unconvincing endpoints include:
1. Simply finding cancers (it must be shown that people benefit from having them found)
2. Preponderance of early-stage cancers (overdiagnosed cases expected to be early stage)

In sum, I believe in the benefits of lung cancer screening for a population very much like the NLST population.
Extending to other groups should be contingent on demonstrating that those groups have a comparable baseline risk of lung cancer. Trials should have a survival outcome

Future work may make it possible to identify such high-risk populations based on another screening test (breath or blood testing), genetic predisposition, or demographic factors (such as ongoing work in Asian females with family history and environmental exposures) 16/17

I sense that today’s discussion has veered away from being a review and towards being a screed. I’ll stop here, and hope that you’ll join me again tomorrow for this month’s final look at locally advanced lung cancer.
17/17