lung cancer screening in smokers

A recent article found that smokers getting low-dose chest CT (LDCT) scans that assessed the volume of nodules had apparently even greater reductions in lung-cancer mortality as well as far fewer false-positive scans than the prior studies assessing nodule diameter, in the NELSON study (see lung cancer volume LDCT nejm2020 in dropbox, or DOI: 10.1056/NEJMoa1911793). I will compare this study with the US National Lung Screening Trial (NLST), on which the US screening recommendations are based

Details:

--13,195 men and 2594 women (men were the primary analysis), aged 50-74, in the Netherlands and Belgium

    --people were either current or former smokers who had quit in past 10 years, who had smoked >15 cigarettes/d for >25 years, or >10 cigarettes/d for >30 years

    --median age of males 58, median smoking history 38 pack-years, 45% were former smokers/55% current

    --cigarettes/d mostly 11-25; duration of smoking mostly 31-45 yrs; age initiating smoking mostly 15-29 yrs

--randomized to 4 CT screenings (at baseline, year 1, year 3, and year 5.5) vs no screening

--among men, screening adherence: 90%

--9% of scans were indeterminate, leading to repeat scans, most often in the baseline scan (20%), then decreasing to 2-7% in subsequent years

--positive LDCTs: referred to pulmonologist if solid nodules with volume >500mm³, pleural-based solid nodule with minimal diameter of >10mm, or solid component in a partial solid nodule with volume >500mm³; or if volume doubling time in subsequent screening <400 dagen (whatever a dagen is...). Also, follow-up scans after 6-8 weeks or 12 months if indeterminate scans (for the full algorithm of nodule management, see their figure S8 in the supplement)

--followup 10 yrs

Results:

among men:

--9.2% had at least one additional imaging study

--overall referral rate to pulmonologists for positive nodules found on LDCT: 2.1%

--overall positive predictive value of screen was 44%

    -- 264 of 22,600 screened participants over all screening rounds (1.2%) had a false positive test

--at 10 yrs of follow-up, in men per 1000 person-years:

    --incidence of lung cancer: 5.58 cases in screening group (341 lung cancers); 4.91 cases  in controls (304 lung cancers), a nonstatistically significant 14% increase with screening, rate ratio 1.14 (0.97-1.33)

        --59% of the cancers found in the screening group were found through screening:

            -- 47% stage 1A, 12% stage 1B; 61% adenoca, 19% squamous cell, 6% small cell [stage 1 lung cancers are largely curable; though some seem to spontaneously regress or at least not progress]

        --41% of cancers in screened group were not detected by screening (ie, interval cancers)

            --72% were stage III or IV; 40% adenoca, 27% squamous cell, 19% small cell

    --cumulative rate ratio for death from lung cancer: 24% less by screening, 0.76 (0.61-0.94), p=0.01, though not much further improvement after year 8

    --all-cause mortality: no difference, 13.93/1000 person-yrs with screening vs 13.76 in control group. cause-specific mortality:

        --lung cancers: total of 370 (160 LDCT group vs 210 in controls)

        --other neoplasms: total of 607 (318 in LDCT and 289 in controls)

        --cardiovascular: total 370 (189 in LDCT and 181 in controls)

--the cumulative incidence of stage 3 or higher lung cancer in the screening group was did not change after the 1st year after randomization, and was a remarkably flat horizontal line from years 1 to 10 (ie, no further accumulation of these cases after the first year: a significant stage shift to less advanced tumors)

--lung cancer mortality: their Figure S6 in the supplement found that the above protocol in NELSON seemed to confer less lung cancer mortality (reduction around 20% after 11 years) vs those who had annual screening (about 13% reduction). [ though it is possible that those getting the annual screenings were different than those on the standard protocol]

--no difference in men or women in rate of lung cancer deaths, comparing either those at 6 years of followup vs those up to 11 years, both overall or in those who would have been eligible in NLST (see below)

among women: 

--rate ratio for lung cancer deaths of 0.67 (0.38-1.14) at 10 years, with values of 0.41 (0.19-0.84) at 8 years and 0.52 (0.28-0.94) at 9 years

sensitivity analyses:

--men 50-54yo (age group not included in the NLST study) had lung cancer mortality rate ratio of 0.85 (0.48-1.50), not statistically significant. the group with most benefit: men 65-69 yo, with a 41% decrease, rate ratio of 0.59 (0.35-0.98)

--50% of the men who were NLST-eligible: rate ratio 0.82 (0.64-1.05), a strong trend to statistical significance

--including all lung cancer deaths (not just the ones with known date of diagnosis, as in the numbers above), for all men:

    --overall rate ratio of 24% reduction with LDCT, RR 0.76 (0.62-0.94), statistically significant

    --among NLST-eligible men, 19% reduction, RR 0.81 (0.63-1.04), strong trend to statistical significance

Commentary:

--lung cancer is leading cause of cancer death worldwide: 18% of cancer deaths, and more than the combo of deaths from breast, colorectal, and cervical cancers in aeras of the world where there are programs to screen for these cancers

    --and lung cancer has a terrible prognosis (one of the reasons it is such a killer): only 15% are alive at 5 years

    --one big issue is that most people present with advanced lesions, hence the focus on early detection

    --prior studies of chest xray (CXR) screening alone did not find significant lung cancer mortality reductions

    --the good news is that smoking prevalence has decreased in western countries, though smoking initiation in youths continues (and, is perhaps exacerbated by the availability of vaping)

--the US National Lung Screening Trial (NLST) was a 3 year study comparing low-dose CT screening annually vs CXR, enrolling 53,454 people aged 55-74 who had history of at least 30 pack-yrs of cigarette smoking and who either were current smokers or had quit in the past 15 years. From 2002-4, and followed 6.5 years (4.5 years after the final screen). NSLT found a 20% decrease in lung cancer mortality and a 6.7% decease in all-cause mortality (see http://gmodestmedblogs.blogspot.com/2014/10/lung-cancer-screening.html )

--the USPSTF, from a modeling study based on NLST, recommended yearly LDCT screening in those aged 55-80yo with smoking history of >30 pack yrs who currently smoke or quit within the past 15 yrs

    --of note, the major difference between these USPSTF recommendations and NLST was further prolonging the screening interval from up to 74yo to 80yo (now exposing smokers to up to 25 LDCTs in those who continue to smoke at least til age 65), as well as increasing the number of annual screens from 3 to 25 based just on mathematical modeling. Interestingly, Medicare, which typically rubberstamps the USPSTF recommendations, lowered the maximum age from 80 to 77: see http://gmodestmedblogs.blogspot.com/2019/04/medicare-and-lung-ct-screening-of.html , which decreased the maximal total number of annual screens down to 22. 

    --there are huge concerns about the much longer LDCT screening interval of the USPSTF recommendations over NLST, the single study leading to the USPSTF recommendations:

        --the 3rd and last year of screening in NLST revealed a somewhat decreasing lung cancer pickup (though not statistically significant). prior to making the expansive USPSTF recommendation, it would have been great to have data showing that the NLST lung cancer pickup rate continued to be high (and LDCT beneficial) with annual screenings for 20-25 years (or at least more than 3 years). but, alas, there are no such data

        --the 20% decrease in lung cancer mortality in NLST translated to a diference of only 62 deaths per 100,000 person-yrs, up to 4.5 years after the screening stopped. would increasing the number of screens per USPSTF find more decreases in the absolute lung cancer mortality benefit?? probably, but would be best to have data....

        --one really big concern i have is the radiation exposure: modeling from other studies (which largely were extrapolated from the general population exposed to radiation) predicted there would be one cancer death created per 2500 people screened by LDCT.  but, smokers likely had lung parenchymal changes already, and the added insult of radiation exposure (multiple hit hypothesis for cancer development) would likely be more carcinogenic than in nonsmokers (for example, controlling for amount of cigarettes smoked, an older study found a 3-fold higher rate of lung cancer in that those who developed COPD: ie, just the changes in lung tissue from COPD seemed to confer higher lung cancer risk)

        --and, in NLST, LDCT, which has an average radiation dose of 1.5 mSv, actually translated into a real average of 8 mSv (about that of a regular chest CT) when one includes the very large number of abnormalities picked up by the LDCT, leading to further imaging (regular CTs, PET scans, etc)

--the major differences between NELSON and NLST include: age 50-74 vs 55-74, timing between the screenings (roughly 2 yrs vs annual), 4 vs 3 screenings, requiring about 15-18 vs 30 pack-yrs smoking, including those who quit up to 10 yrs vs 15ys prior to screening, and followup of 10 vs 6.5 yrs

    --findings favoring the NELSON approach over NLST:

        --the percentage of patients with a positive test on screening was 2.1% with NELSON's volume-based assessment vs 24% with NLST's diameter-based one

        --the benefis in NELSON were even better than NLST, though the former had larger intervals between screenings ( ie, their 2-year screening interval was twice NLST, providing perhaps the potential to decrease the total number of screenings/exposure to radiation)

        --the positive predictive value in NELSON was 44% vs 4% in NLST

        --and, modeling studies suggest that the overdiagnosis rates in NELSON was 9% vs 19% in NLST

And for all LDCT screening, there may be a really unfortunate other adverse effect: patients with normal LDCTs may think they are fine and it is okay to continue smoking (really missing the message that they are more likely to die from cardiovascular disease or some other non-lung cancers. And, in the NELSON study, non-lung cancer mortalities were in fact 65% more likely overall than lung cancer ones and there were equal numbers of lung cancer and cardiovascular deaths!!! (in fact lung cancer caused only about 20-25% of the total deaths....)

So, the concern is that LDCT screening focuses solely on lung cancer from cigarette smoking, and this may dwarf and distort the importance of other very serious smoking-related problems in patients). see http://gmodestmedblogs.blogspot.com/2015/08/another-downside-of-lung-cancer-ct.html .

    --also a real-world study of LDCT per the USPSTF guidelines done in the VA, found a really huge, 97.5% false positive rate (with the likely attendant increased anxiety, stress, etc, as well as ensuing radiation exposure): http://gmodestmedblogs.blogspot.com/2017/02/lung-cancer-screening-results-at-va.html

    --there are quite legitimate concerns about the entry criteria for NLST and USPSTF recommendations: there should have been a better risk-based oriented screening, since their 30 pack-yr cutpoint excludes some high-risk people and includes some low-risk ones. eg see  http://gmodestmedblogs.blogspot.com/2018/01/risk-based-low-dose-ct-screening-in.html and http://gmodestmedblogs.blogspot.com/2016/07/lung-cancer-screening-for-smokers.html, 2 blogs suggesting that risk-based screening (which does include COPD symptoms) was more accurate in choosing people to screen with LDCT than just the pack-yrs, per the Natl Lung Screening Trial and endorsed by USPSTF. NELSON also did not use a risk-based approach

--though this NELSON study focused on males, their results from the small subset of women found that the effects of screening were even more profound in women, as has been found in other studies (including in NLST)

-------------------------------------------------------------

--Extended analysis of the NLST at 11.3 years, about 9 years after the last LDCT screen (see lung cancer CT screen NLST fu jthoraconc2019 in dropbox, or doi.org/10.1016/j.jtho.2019.05.044):

    --1701 cancers found in the LDCT group vs 1681 in the CXR group. non-significant difference

    --lung cancer deaths: 1147 with LDCT vs 1236 with CXR, 8% reduction, RR 0.92 (0.85-1.00)

       --difference in lung cancer deaths between the LDCT and CXR groups was 3.3/1000, with number-needed-to screen being 303 (essentially the same as in the original NLST analysis finding NNS of 320)

        --dilution-adjusted lung cancer mortality was 11% less, RR 0.89 (0.80-0.997) [since the NLST stopped after 3 years and with prolonged followup, they wanted to avoid the dilution effect: new cancers might develop after the screenings stopped, but that these would likely occur in both groups equivalently, and would dilute/distort the true effects of the intervention itself]. Note: in the original study, there was  a 20% reduction [this decreasing mortality benefit might signify the advantage of more screening over a larger time. but is this true?? is there a difference between those continuing to smoke vs having stopped 10-15 years before? should we screen annually vs every 2 years. or every 5 years. or is there a way to stratify this interval by risk???]

    -- overall mortality: no difference, 5253 with LDCT and 5366 with CXR, difference of 4.2/1000 but confidence interval of -2.6 to 10.9

    --and, the graphs of lung cancer mortality showed increasing benefit of LDCT for about 3-4 years, then significant narrowing of the differences of between LDCT and CXR thereafter

 ---------------------------------------------------------------

--one concern in screening tests for cancer is lead-time bias (just picking up cancers earlier, leading to an apparent but illusory benefit).

    -- mathematical modeling (see lung cancer screening leadtime bias jepidcommhlth2015 in dropbox, or ten Haaf K. J Epidemiol Community Health 2015; 69: 1035), suggests that lead-time bias might be in the 9-12 year range for lung cancers. all of this meaning that we should have better data from NELSON with a few more years of followup

--a few limitations of the NELSON study:

    --we do not know the radiation exposure for the volume-based LDCT (??different from the NSLT with their diameter-based algorithm)

    --we do not know the overall average radiation exposure in NELSON, combining the initial LDCT plus whatever imaging ensues from indeterminate or positive initial scans (as above, was actually very high in NSLT)

    --there was no evaluation of lung cancer mortality stratified by the intensity of cigarette smoking or by those who quit vs continued smoking

so, an interesting study, suggesting a few conclusions:

--volume-based nodule assessment for smokers seems to be better: fewer false positives, probably overall less radiation, probably fewer referrals to pulmonary/procedures (not able to quantitate from the article, but from their numbers seems to be fewer), though both studies show lower lung cancer mortality.

--and, they confirmed signifcant lung cancer reductions: 24% in men and 33% in women, associated with even less aggressive LDCT screening (ie, not annually)

--as per blogs noted above, a risk-based approach may be more appropriate

--but the real major concerns to me are:

    --the big issue with lung cancer screening is stopping the carcinogens. mostly smoking, but also occupational and environmental ones. and, my guess is that funding public health initiatives and vigorous smoking cessation programs would be better for the smokers and cost a lot less than LDCTs. And perhaps a public health approach to decrease smoking initiation, specifically targetting those <30yo, the age when most start smoking, to decrease access to vaping products/cigarette access, would likely help a lot. And it would be great to have increased regulations and large-scale public health initiatives to decrease the environmental and occupational exposures (smokers have a much higher risk of lung cancer with these added exposures)

    --and the other big concern is  that patients (and we clinicians) may focus so much on lung cancer by doing LDCTs that the overwhelming other morbidities and mortalities from smoking are diminished or neglected. And a normal LDCT may give a false sense of security...

geoff​

If you would like to be on the regular email list for upcoming blogs, please contact me at gmodest@uphams.org

to get access to all of the blogs:

1. go to http://gmodestmedblogs.blogspot.com/ to see them in reverse chronological order

2. click on 3 parallel lines top left, if you want to see blogs by category, then click on "labels" and choose a category​

3. or you can just click on the magnifying glass on top right, then  type in a name in the search box and get all the blogs with that name in them

please feel free to circulate this to others. also, if you send me their emails, i can add them to the list