Risk-based low-dose CT screening in smokers???

​​Yet another analysis suggests that it is better to do risk-based low-dose CT screening of smokers instead of one based on the USPSTF (US Preventive Services Task Force) eligibility criteria (see doi:10.7326/M17-2067).

Background:

-- The NLST (Natl Lung Screening Trial) was published in 2011 and assessed 26,722 smokers between age 55-74 who had smoked at least 30 pack-years and, if former smokers, had quit within the prior 15 years, and found that doing 3 annual low-dose CT (LDCT) screenings was better than chest xrays, with a reduction of 62 lung cancer deaths/100,000 person-years, a 20% relative reduction (see doi:10.1056/NEJMoa1102873 )

--as a result of this single study, the USPSTF recommended annual screening of all smokers aged 55-80 who had smoked 30 pack-years and currently smoked or quit within the past 15 years. smokers would be screened annually until they reach the 15-year mark for not smoking, or until age 80 (ie, a potential of 25 annual screens for those continuing to smoke)

--Medicare in 2015 down-graded this a bit, adopting the same USPSTF criteria, but decreasing the age range to 55-77 (ie, a max of 22 annual screens).

Details of this study:

--the researchers compared nationally representative patient groups from the National Health Interview Survey from 2005, 2010, and 2015, including 5-7000 people from each study who were ever-smokers aged 50-80

--they calculated individual 5-year risks for lung cancer incidence and deaths using 2 tools: the Lung Cancer Risk Assessment Tool and the Lung Cancer Death Risk Assessment Tool

--they then estimated the number of these smokers who would be eligible for LDCT per the USPSTF guidelines vs per the individual 5-year risk  for lung cancer of at least 1.5%, 2%, or 2.5%

--then they estimated the number of lung cancer deaths potentially preventable by these screenings, and the number of US smokers in subgroups eligible and ineligible for screening per USPSTF

Results:

--for screening:

   --using the USPSTF criteria:

      --from 2005-2010: number eligible for screening increased from 8.7 million (24.6% of ever-smokers aged 50-80) to 9.5 million (22.8%)

      --from 2010-2015: number eligible for screening decreased 1.5 million to 8.0 million (18.4% of ever-smokers aged 50-80)

   --using the 1.5% 5-year lung cancer risk cutpoint: 0.7 million fewer US smokers would be screened from 2010-2015

   --using the 2.0% 5-year lung cancer risk cutpoint: 0.8 million fewer US smokers would be screened from 2010-2015

   --using the 2.5% 5-year lung cancer risk cutpoint: 0.8 million fewer US smokers would be screened from 2010-2015

--projected US lung cancer preventable deaths by screening, from 2010 to 2015:

   --USPSTF: 8122 fewer deaths averted in 2015: 6.4 percentage point fewer preventable deaths averted by following these guidelines

   --1.5% 5-year lung cancer risk cutpoint: approx. 6000 fewer deaths averted, 2.6 percentage point fewer preventable deaths

   --2.0% 5-year lung cancer risk cutpoint: approx. 6000 fewer deaths averted, 3.2 percentage point fewer preventable deaths

   --2.5% 5-year lung cancer risk cutpoint: approx. 6000 fewer deaths averted​, 3.7 percentage point fewer preventable deaths

--as a perspective, using the >2.5% cutpoint: in 2005, this individual risk group would have prevented 2617 more deaths than USPSTF; in 2015 the difference would be 5115 patients

Commentary:

--one positive development is that smokers overall are smoking less than before, and younger smokers smoke less than older smokers (in 2015 smoking prevalence in those 80 yowas 47%, with mean of 16.9 cigarettes/d vs in those 40 yo the prevalence was 34% and mean of 11.7 cigarettes/d, p<0.001 for trend). So, fewer people are reaching the 30 pack-year cutpoint of the NLST or the USPSTF recommendations.  And this probably explains the decrease in eligible smokers for LDCT screening. [the smoking prevalence numbers here seem high to me, but are quoted in the article without a clear reference. but other sources confirm at least the main points: fewer people are smoking now and the quantity smoked has decreased pretty markedly]

--this study found that of the smokers smoking enough to be eligible for screening by any of the above criteria, twice as many would not be screened if sticking to USPSTF (ie, a relatively larger % would qualify to be screened if using the individual criteria). And, assuming that the actual risk is as per the calculators, using the individual risk-based approach would reach many more people (esp those smoking in the 20-29 pack-years range) and lead to lower lung cancer mortality (comparing the 2010-2015 numbers, the decrease in the number eligible for screening was ½ as much as with the USPSTF criteria)

--and, even a post-hoc analysis of the NLST itself found that there was a pretty dramatic difference in lung-cancer deaths by lung cancer risk quintile (from 0.2/10,000 person-years in the lowest risk group to 12.0/10,000 person-yrs in the highest, p=0.01 for trend), associated with a dramatic decrease in false positive results along this risk gradient (1648/10,000 person-yrs in the lowest risk group, down to 65/10,000 person-yrs in the highest), leading to number-needed to screen going from 5276 down to 161 in the highest risk group. The 60% of people in the highest 3 quintiles accounted for 88% of the screening-prevented lung cancer deaths and for 64% of those with false positive results.  And the 20% at lowest risk accounted for only 1% of the prevented lung cancer deaths (see Kovalchik SA. N Engl J Med 2013; 369: 245). In this study they developed their own absolute risk-prediction model from the NLST’s control group (CXR only), which was externally validated with  in the PLCO (Prostate, Lung, Colorectal, and, Ovarian cancer screening trial)

--a prior blog reviewed a different study of risk-based screening; see http://gmodestmedblogs.blogspot.com/2016/07/lung-cancer-screening-for-smokers.html , a study which assessed 3 databases (CXR-only wing of NLST, ever-smokers control group in PLCO, and NHIS), finding that some patients not qualifying for the recommended LDCT (and would not be screened per the USPSTF) are at higher risk than some who would qualify: ieunderscreen some but overscreen others

--and, a cost-effectiveness analysis of NLST (see doi:10.7326/M17-1401) found that the mortality benefits of a risk-based approach were considerable, though the cost-effectiveness gap was not so large:

   --the incremental lung cancer mortality benefits during the first 7 years ranged from 1.2-9.5 lung cancer deaths prevented/10,000 person-yrs, depending on the decile of their risk category

   --the gradient of life-years was less remarkable, since the higher risk group overall was older, had higher smoking exposure, were more likely to have COPD and were more likely to have invasive testing (all increasing their baseline mortality risk).

   --the incremental cost-effectiveness ratios (ICERs) varied from $75,000 per quality-adjusted life-years (QALY) in the lowest decile of risk to $53,000 per QALY in the highest (again, part of the issue here is, as noted above, those in the highest-risk group were sicker and older and had more invasive/costly testing after a positive LDCT screen).

  --of note, this analysis was for NLST, with only 3 annual screenings, and does not apply to following the potentially many more LDCTs per USPSTF or per Medicare.

so, my concerns about LDCT, which are elaborated in several prior blogs (see above/below), include:

--the unsupported extension of the 3-year NLST trial LCT screenigs to potentially 25 years by USPSTF if the patient continued to smoke (downgraded to 22 years by Medicare, topping off at age 77 instead of 80, which, by the way, is the only time that Medicare has deviated from the USPSTF that I have seen, since Medicare is legislatively obligated to follow USPSTF recommendations). And both of these age ranges are more than NLST, which was 55-74. see http://gmodestmedblogs.blogspot.com/2019/04/medicare-and-lung-ct-screening-of.html for the Medicare recommendations

--the results of NSLT found an impressive 20% decrease in lung cancer mortality, but this translated to a difference of only 62 deaths per 100,000 person-years

--the 3-year data on lung cancer pickups in NLST was actually decreasing by the 3^rdyear, which also goes against the extension to potentially so many more years.

--the average dose of radiation in LDCT screening (1.5 mSv) turned out to be actually the same as with regular CT screening (8mSv), when one adds in the commonly needed extra imaging (full CTs, PET scans, etc) for the many positive (and markedly high rate of false-positive) LDCT scans

--this dose of radiation is modeled to create one cancer death per 2500 people screened.  This last point does bring up one concern of mine in this modeling: there clearly are individual differences in the effects of smoking on individuals' lungs; at the same quantity of cigarette smoking, some people develop severe parenchymal changes (COPD etc) while others do not. And this seems to be important: those with COPD, for example, have a 3-fold increased risk of lung cancer over those who do not (again, correcting for the amount of cigarette smoked). And, it may well be that those who already have significantly smoking-damaged lungs are at a much higher risk of radiation-induced cancer.... (ie, those who have worse lungs from the effects of smoking, who then get repeated LDCTs over many years, may have really dramatically increased risk of radiation-induced lung cancer than the 1 in 2500 in the above model, which was mostly based on population data from atomic bomb explosions and some medical imaging studies, and probably mostly from people with baseline normal lungs.....)

--the focus on lung cancer in smokers may dangerously undercut probably the most important smoking-related morbidity and mortality, atherosclerotic cardiovascular disease, creating the impression (to patients, and perhaps somewhat to clinicians) of “all is great” if the LDCT were normal

 --but, these LDCT recommendations really put us in primary care in a bind: these are strong recommendations from a pretty conservative group overall (USPSTF, then Medicare), and even if one feels that they reflect not only overstated extrapolated results from a single study (NLST) and might really undercut the real message that smoking is really terrible and especially leads to heart disease much more than lung cancer, we are really in a bind: if a smoker meeting USPSTF criteria for LDCT does get lung cancer, we are really open to a malpractice case.  even though we clinicians, so far, seem to be referring only about 2% of USPSTF-eligible patients for LDCT.....

see http://gmodestmedblogs.blogspot.com/2014/10/uspstf-lung-cancer-screening-revisited.html   for a review of the USPSTF recommendations, given a “B” grade (which means they recommend it with high certainty of moderate net benefit or moderate certainty of moderate to substantial benefit), as of December 2013, which, by the way, was finalized after the post-hoc analysis of NLST showing a risk-based approach was much better (see above)

see http://gmodestmedblogs.blogspot.com/2019/04/medicare-and-lung-ct-screening-of.html for the blog on Medicare’s recommendation for LDCT, released after the USPSTF one

see http://gmodestmedblogs.blogspot.com/2017/02/lung-cancer-screening-results-at-va.html , a blog about implementation of LDCT in the VA,  finding a huge 97.5% false positive rate, further critiquing the recommendations, and reinforcing the need to have studies done in different settings before developing a national policy

see http://gmodestmedblogs.blogspot.com/2016/03/need-annual-low-dose-chest-cts.html  , a blog reviewing a retrospective cohort analysis of the NLST, finding that an initial negative LDCT actually seemed to have a much lower subsequent incidence of lung cancer and that annual screening may not be necessary

See https://dceg.cancer.gov/tools/risk-assessment/lcrisks for lung cancer risk assessment tools

Geoffrey Modest, MD

to be added to the email list: contact me at gmodest@uphams.org