Man oh man, lung cancer screening. I’ve put off developing an opinion on this for about six years now as I have waited for the data to accrue (or rather as I have tried not to drown in residency), but the USPSTF was much more proactive. All major payors have covered lung cancer screening in high risk patients on the basis of the National Lung Cancer Screening Trial (NLST) since 2014. But anyone who’s had an outpatient primary care patient recently will probably confess that lung cancer screening isn’t on the same mental check list as colonoscopy, mammography, and cervical cancer screening. It just has not yet penetrated primary care to the degree to which a screening intervention needs in order to be effective. Worries abound about the difficulty in following suspicious nodules with serial CT scans, and talking patients into yet another screening test has its own challenges.
So I was thrilled to see early reports from the VA’s large-scale implementation of lung cancer screening (LCS) in their population. More a quality improvement project than a study per se, this paper by Linda Kinsinger and colleagues gives us some important insights about the challenges of developing a comprehensive screening program. So, should we be leading the charge to get everyone into the CT scanner, or was early excitement about screening overblown? We’re going to talk about that… tomorrow. But in the name of presenting information in approachable chunks, I’m going to pare back a bit and give you the set up today in preparation for tomorrow’s punchline.
The NLST was published in 2011 to much fanfare. This study enrolled 53,454 current or former smokers with at least 30 pack-years to their name to annual low-dose CT versus chest radiography. How low-dose was it? Great question! When in doubt, ask the ACR. That nifty chart is my favorite resource for remembering that it takes about 70 chest radiographs to equal one CT chest, but only one KUB to deliver the same amount. You’re welcome. As it turns out, one low-dose chest CT delivers just over 20% of the radiation of a standard chest CT. So, not terrible.
Anyway, it worked. More than 90% of patients enrolled were adherent to their respective screening protocols. Over a median 6.5 year follow up, nodules requiring further intervention were found in 24.2% of low-dose CT versus 6.9% of radiography patients. This higher rate turns out to be a very good thing, as patients undergoing CT enjoyed a relative risk reduction for lung cancer death of 20.0% (95% CI, 6.8 to 26.7; P=0.004). This gives us a number needed to screen (NNS) of 320, which the astute will note puts the absolute risk reduction (ARR) at 0.31%. It’s tough to find an apples-to-apples comparison here — colonoscopy, for example, has a NNS of ~20 for detection of advanced adenoma, but no firm data with regard to mortality benefit. Breast cancer screening is controversial, but one (now somewhat outdated) Cochrane review puts the ARR at 0.05%, with a NNS of 2000. So this is certainly comparable in its benefits to things we are already doing.
But if that’s not doing it for you, try this on: there was also an all cause mortality benefit, relative risk reduction 6.7% (95% CI, 1.2 to 13.6). The authors wisely note that this disappears when excluding lung cancer deaths (i.e. the lung cancer mortality benefit drives the overall mortality benefit). Nonetheless, moving the needle on all cause mortality is no mean feat.
So the study wasn’t perfect. The follow up period was not long enough to detect any secondary cancers as a result of excess radiation exposure, for one thing. For another, possible harms of screening were likely underestimated by comparing to radiographic screening rather than no screening — rates of complication were similar between the two screening groups, but would almost certainly have been lower in a “no intervention” control. Given that we don’t actually offer radiography for screening, perhaps this would have been a more meaningful comparison. One letter to the editor does the math for us: “CT screening would also result in 231 people with at least 1 false positive result, 22 additional invasive procedures, 18 additional surgeries, and 6 nonfatal complications resulting from these procedures…” compared to these patients never having been exposed to screening.
Oh, and the study includes this gem: “Since there was no standardized, scientifically validated approach to the evaluation of nodules, trial radiologists developed guidelines for diagnostic follow-up, but no specific evaluation approach was mandated.” That translates into slide 5 of this monstrosity, which serves as the guidelines used in the VA implementation we’re looking at tomorrow. So in the end, this was a trial of a largely made-up screening intervention. And yet: it worked. Pretty damn well.
So We Ran With It
The USPSTF did what the USPSTF does, and correctly noted that the benefits of the study (decreased mortality) were likely to outweigh any harms. After all, what possible complications could outweigh an all-cause mortality benefit? They gave it a grade B recommendation in December of 2013. This had some consequences. First of all, the affordable care act mandated that all private insurance cover, without coinsurance, any USPSTF preventative service with a grade A or B rating. So, this bad boy got paid for. Medicare followed suit shortly thereafter, after initially signaling reluctance to approve the intervention on the basis of this single study, and paid for it starting in December 2014. So, here we are.
The original trial made no reference to the cost-effectiveness of its intervention, but it was not long until someone took the opportunity. They come up with $81,000 per QALY gained and $52,000 per life-year gained, which is on the right side of borderline cost-effective. I tried to find a decent list of cost per QALY for some other interventions, but it’s curiously hard to find. Many people use a threshold of either $50,000 or $100,000, the difference between which should tell you that this is an arbitrary decision based on how much we feel comfortable spending on healthcare. This particular analysis assumed an average of one follow up CT scan per positive screening examination, and a rate of initial positive screens similar to that in the NLST. A few other estimates exist, and discussion of their relative validity is above my pay grade.
So: we have a screening method that worked well in a clinical trial setting and has a mortality benefit as good or better than the majority of our existing screening methods. We also have outstanding questions about its long-term harms, and a complete lack of certainty as to what happens when you take an intervention with an adherence rate of 90% and turn it loose on the world. Hopefully our good friends at the VA can tell us something. Tune in tomorrow to find out!