Latest / This Week In Medicine / Annals of Internal Medicine December 2026 Eligibility and Prognostic Performance of Smoking Duration–Based Versus Pack-Year–Based U.S. National Lung Cancer Screening Criteria Across Racial and Ethnic Groups

Transcript

Annals of Internal Medicine December 2026 Eligibility and Prognostic Performance of Smoking Duration–Based Versus Pack-Year–Based U.S. National Lung Cancer Screening Criteria Across Racial and Ethnic Groups · This Week In Medicine

• 0:00
  [SPEAKER_00]: Welcome back to The Deep Dive.
• 0:01
  [SPEAKER_00]: Today, we are jumping straight into a really crucial conversation around lung cancer screening.
• 0:07
  [SPEAKER_00]: I mean, this is a matter of life and death, and the metrics we're using right now to decide who gets screened.
• 0:13
  [SPEAKER_00]: Well, they might be systematically failing the very people who need help the most.
• 0:17
  [SPEAKER_01]: that's exactly right.
• 0:19
  [SPEAKER_01]: And if you're not familiar with how this works, the standards in the US come from the preventive services task force or the USPSTF.
• 0:26
  [SPEAKER_01]: And their whole decision process has traditionally been based on this one single metric.
• 0:31
  [SPEAKER_01]: Yeah, the pack here.
• 0:33
  [SPEAKER_00]: Okay, let's make sure we're all on the same page with pack here because it's it's absolutely central to this whole thing.
• 0:37
  [SPEAKER_01]: It really is.
• 0:38
  [SPEAKER_01]: A pack here is a pretty simple calculation.
• 0:40
  [SPEAKER_01]: It's the number of packs of cigarettes you smoke per day on average multiplied by the number of years you've
• 0:47
  [SPEAKER_00]: So like one pack a day for 20 years is 20 pack yours.
• 0:50
  [SPEAKER_01]: Exactly.
• 0:51
  [SPEAKER_01]: Or, you know, two packs a day for 15 years gets you to 30 pack yours, simple math.
• 0:55
  [SPEAKER_00]: great.
• 0:56
  [SPEAKER_00]: So it's a mix of both how much you smoked and for how long.
• 0:59
  [SPEAKER_00]: And for a long time, that bar was set pretty high, which meant a lot of smokers just weren't eligible.
• 1:03
  [SPEAKER_01]: Correct.
• 1:04
  [SPEAKER_01]: And the USPSTF knew that.
• 1:06
  [SPEAKER_01]: They recognized the old criteria was too narrow.
• 1:08
  [SPEAKER_01]: So they made a pretty big move in 2021.
• 1:11
  [SPEAKER_01]: They lowered the minimum requirement from 30 pack years down to 20 to try and expand access.
• 1:17
  [SPEAKER_01]: And that was, you know, a very positive step.
• 1:19
  [SPEAKER_00]: but and this is the core of our deep dive today.
• 1:22
  [SPEAKER_00]: Here's the problem.
• 1:23
  [SPEAKER_00]: Even after they expanded the criteria, we are still seeing these dramatic persistent disparities.
• 1:31
  [SPEAKER_00]: The data just shows that underrepresented racial and ethnic minorities are facing these systemic barriers.
• 1:36
  [SPEAKER_00]: They're less likely to be eligible than their white counterparts, even when they're risk of lung cancers the same, or maybe even higher.
• 1:43
  [SPEAKER_01]: So that brings us to the mission for today.
• 1:45
  [SPEAKER_01]: We're going to unpack this really compelling new study from the Annals of Internal Medicine.
• 1:50
  [SPEAKER_01]: And it basically asks a fundamental question.
• 1:52
  [SPEAKER_01]: Is the whole idea of pack years inherently flawed as a way to measure risk equitably?
• 1:56
  [SPEAKER_00]: And what happens if we just toss it out?
• 1:59
  [SPEAKER_00]: What if we use something simpler, like just how long you've smoked, or something way more complex, like a high-tech risk model?
• 2:06
  [SPEAKER_01]: We want to find out which approach actually detects the cancer effectively.
• 2:10
  [SPEAKER_01]: But, and this is the critical part, which one also closes that eligibility gap for groups like African Americans and Latinos.
• 2:16
  [SPEAKER_00]: Okay, so we're comparing the current standard, the USPSTF-2021 criteria against two other options.
• 2:22
  [SPEAKER_01]: Yep, a simple duration-based model, and then a very sophisticated risk-based model.
• 2:27
  [SPEAKER_01]: It's all about finding the fairest way to provide access to care.
• 2:31
  [SPEAKER_00]: All right.
• 2:31
  [SPEAKER_00]: Let's dig in.
• 2:32
  [SPEAKER_00]: To answer question about equity, the researcher is obviously needed a huge diverse data set.
• 2:36
  [SPEAKER_00]: So, where did this data come from?
• 2:39
  [SPEAKER_01]: This is really the study's strength.
• 2:40
  [SPEAKER_01]: It's a prospective population based analysis, and it draws from the multi-ethnic cohort study or MEC.
• 2:47
  [SPEAKER_01]: And this wasn't just some small sample.
• 2:48
  [SPEAKER_01]: It was linked directly to the ZAKR registries.
• 2:51
  [SPEAKER_01]: That's the surveillance epidemiology and end results program, which tracks cancer outcomes nationwide.
• 2:56
  [SPEAKER_00]: So, a really solid foundation and who is in this cohort.
• 2:59
  [SPEAKER_01]: It was a massive pool of people.
• 3:02
  [SPEAKER_01]: Over 100,000 adults, 105,261 to be exact, all aged between 45 and 75, and all with a history of smoking.
• 3:12
  [SPEAKER_01]: And, crucially, they were recruited back in the mid-90s, mostly in California and Hawaii, that part is important.
• 3:19
  [SPEAKER_01]: Well, that geographical scope gave the researchers the diversity they needed to actually study these racial and ethnic gaps in eligibility.
• 3:28
  [SPEAKER_00]: Right, so it wasn't just a study focused on one single demographic.
• 3:31
  [SPEAKER_00]: Okay, so the researchers were comparing these different screening criteria.
• 3:34
  [SPEAKER_00]: What were they measuring?
• 3:35
  [SPEAKER_00]: Exactly.
• 3:35
  [SPEAKER_00]: I know they looked at eligibility, but what about performance?
• 3:38
  [SPEAKER_01]: Good question.
• 3:39
  [SPEAKER_01]: So prognostic performance really boils down to two vital statistics for any kind of screening program.
• 3:46
  [SPEAKER_01]: Sensitivity and specificity.
• 3:47
  [SPEAKER_00]: Okay, break those down for us.
• 3:48
  [SPEAKER_01]: Sensitivity is pretty simple.
• 3:49
  [SPEAKER_01]: It's the test ability to correctly find the people who actually have lung cancer within a certain time frame.
• 3:54
  [SPEAKER_01]: in this study, it was six years.
• 3:57
  [SPEAKER_01]: We call those the true positives.
• 3:58
  [SPEAKER_00]: So it's about catching the people who are really sick.
• 4:01
  [SPEAKER_01]: Exactly.
• 4:01
  [SPEAKER_01]: But you can't just have sensitivity or you just screen everybody.
• 4:05
  [SPEAKER_01]: That's where specificity comes in.
• 4:07
  [SPEAKER_01]: Specificity is the ability to correctly identify the healthy people to not flag them.
• 4:13
  [SPEAKER_01]: Failures there are called false positives, telling a healthy person they might be sick.
• 4:17
  [SPEAKER_00]: And a high rate of false positives.
• 4:19
  [SPEAKER_00]: I mean, that's not just a small problem.
• 4:21
  [SPEAKER_00]: That has real consequences, doesn't it?
• 4:23
  [SPEAKER_01]: Oh, absolutely.
• 4:24
  [SPEAKER_01]: It leads to unnecessary, often invasive, follow-up tests.
• 4:28
  [SPEAKER_01]: It causes huge anxiety for patients, drives up health care costs.
• 4:33
  [SPEAKER_01]: And you know, there's always a risk of complications from things like a biopsy you never even needed.
• 4:37
  [SPEAKER_01]: It's the constant balancing act of screening.
• 4:39
  [SPEAKER_00]: Finding that sweet spot.
• 4:40
  [SPEAKER_01]: Right.
• 4:41
  [SPEAKER_01]: Catching the true cases while minimizing the harm from all the false alarms.
• 4:44
  [SPEAKER_00]: makes sense.
• 4:45
  [SPEAKER_00]: So before they tested the alternatives, what did the researchers find was the baseline?
• 4:50
  [SPEAKER_00]: What was the problem they were starting with?
• 4:51
  [SPEAKER_01]: The baseline, using the current USPSTF-2021 rules, the expanded criteria we mentioned, was that only 24% of this entire high risk cohort was even eligible for screening.
• 5:02
  [SPEAKER_00]: Only 24% so three quarters of this diverse group of smokers
• 5:08
  [SPEAKER_01]: That's the problem in a nutshell.
• 5:10
  [SPEAKER_00]: Okay, so that brings us to the first alternative they tested.
• 5:13
  [SPEAKER_00]: The idea of just getting rid of the pack your math altogether and using a simpler metric, how long a person has smoked.
• 5:20
  [SPEAKER_00]: The duration criteria.
• 5:22
  [SPEAKER_01]: Exactly.
• 5:23
  [SPEAKER_01]: They play with a few different thresholds and they found that using a 30-year smoking duration gave them the most useful comparison.
• 5:30
  [SPEAKER_01]: At that 30-year cutoff, the overall eligibility rate was about 27.5%, which was pretty close to the 24% from the current standard.
• 5:39
  [SPEAKER_01]: So there's screening a similar slice of the population.
• 5:41
  [SPEAKER_00]: Okay, a small bump in the total number of people eligible.
• 5:45
  [SPEAKER_00]: But this is where it gets really interesting because the goal wasn't just more people, it was fairer access.
• 5:49
  [SPEAKER_00]: So what did this simple 30-year rule do to those eligibility gaps between different racial and ethnic groups?
• 5:55
  [SPEAKER_01]: The change is, but it's stunning, and it's really the most powerful finding in the whole study.
• 6:01
  [SPEAKER_01]: Compared to the current standard, the 30-year duration criteria dramatically reduce the eligibility gaps for every group relative to whites.
• 6:09
  [SPEAKER_01]: We have to look at the numbers.
• 6:10
  [SPEAKER_00]: Yes, please, let's get into the specifics because this is the detail you really need to grasp.
• 6:14
  [SPEAKER_01]: OK, so under the standard USPST of 2021 criteria, 30.2% of white participants were eligible for screening.
• 6:23
  [SPEAKER_01]: But for African Americans, that number dropped way down to 21.4%.
• 6:27
  [SPEAKER_01]: And for Latinos, it was even lower, just 15.7% eligible.
• 6:32
  [SPEAKER_01]: That's the gap.
• 6:33
  [SPEAKER_00]: so you can see it right there.
• 6:35
  [SPEAKER_00]: African Americans and Latinos were being disproportionately screened out by that pack your calculation.
• 6:40
  [SPEAKER_00]: So what happens when you just switch to a 30 year smoking history?
• 6:44
  [SPEAKER_01]: The numbers completely flip.
• 6:45
  [SPEAKER_01]: For African Americans, it's just a dramatic change.
• 6:48
  [SPEAKER_01]: Under this new duration model, the white eligibility is about 28.8% so a tiny dip.
• 6:53
  [SPEAKER_01]: But for African Americans, the eligibility rate rockets up from that 21.4% to 30.4%.
• 6:58
  [SPEAKER_00]: Wait a bit, hold on a second.
• 7:00
  [SPEAKER_00]: Did you just say the rate for African Americans surpassed the rate per whites just by changing that one metric?
• 7:05
  [SPEAKER_01]: That is exactly what the data shows under the simple duration rule.
• 7:08
  [SPEAKER_01]: African American eligibility actually exceeded white eligibility.
• 7:11
  [SPEAKER_01]: It didn't just close the gap.
• 7:12
  [SPEAKER_01]: It erased it and even reversed it slightly.
• 7:14
  [SPEAKER_00]: That is just a wild statistical correction from what seems like such a simple policy change.
• 7:20
  [SPEAKER_00]: Why does that happen?
• 7:21
  [SPEAKER_00]: Why does duration work so much better than pack years for these groups?
• 7:26
  [SPEAKER_01]: So the study doesn't pinpoint the exact reason, but if you look at the broader picture, it strongly suggests that the smoking patterns for African-American and Latino populations are just different.
• 7:38
  [SPEAKER_01]: They don't align with how pack years measure risk.
• 7:40
  [SPEAKER_00]: You mean they might smoke for a really long time, but maybe less intensely fewer cigarettes per day.
• 7:46
  [SPEAKER_01]: That's the likely explanation.
• 7:47
  [SPEAKER_01]: Yeah.
• 7:48
  [SPEAKER_01]: They might have a very long duration of exposure, which is a major risk factor.
• 7:51
  [SPEAKER_01]: But their pack year number never climbs above that 20-cack year cutoff.
• 7:55
  [SPEAKER_01]: The duration metric actually captures their real world risk much better.
• 7:59
  [SPEAKER_00]: And it was a similar story for Latinos, right?
• 8:01
  [SPEAKER_00]: Where the gap was the biggest to start with.
• 8:03
  [SPEAKER_01]: Yes, a huge effect there as well.
• 8:05
  [SPEAKER_01]: Their eligibility jumped from that really low 15.7% all the way up to 25.1% with the duration model.
• 8:11
  [SPEAKER_01]: It's a massive increase in access for a group that was being left behind.
• 8:15
  [SPEAKER_00]: So on the equity front, it seems like a no-brainer.
• 8:18
  [SPEAKER_00]: The duration model is a clear winner over the current standard.
• 8:21
  [SPEAKER_00]: But you mentioned a trade-off, what of the quality of the screening?
• 8:25
  [SPEAKER_01]: and this is the nuance that policymakers always have to grapple with.
• 8:29
  [SPEAKER_01]: The duration model did improve sensitivity for detecting lung cancer across all races.
• 8:34
  [SPEAKER_00]: Okay, so it caught more true cancer cases, that's good.
• 8:36
  [SPEAKER_01]: That's very good.
• 8:37
  [SPEAKER_01]: But, specificity went down at the same time.
• 8:40
  [SPEAKER_00]: Ah, so, better at finding the sick people, but also more likely to flag a healthy person for more tests.
• 8:47
  [SPEAKER_01]: That's the dilemma.
• 8:49
  [SPEAKER_01]: You get more true positives, but you also get more false positives.
• 8:52
  [SPEAKER_01]: The risk of overscreening goes up.
• 8:54
  [SPEAKER_01]: It's a critical point.
• 8:56
  [SPEAKER_01]: But when you weigh it against the enormous equity gains, you could argue that tradeoff is worth it as long as the false positive rate is manageable.
• 9:03
  [SPEAKER_00]: Okay, that sets up the perfect transition to the third option they looked at.
• 9:07
  [SPEAKER_00]: If duration had this tradeoff, did they try to find the best of both worlds with a pure data model?
• 9:12
  [SPEAKER_00]: Some kind of complex algorithm.
• 9:15
  [SPEAKER_01]: They did.
• 9:15
  [SPEAKER_01]: They tested a very sophisticated risk-based model, it's called the Re-Calibrated PLC-Om 2012 update.
• 9:23
  [SPEAKER_01]: And this model doesn't just look at smoking, it integrates a whole bunch of variables, age, BMI, family history, and other factors to calculate a very precise six-year risk score.
• 9:33
  [SPEAKER_01]: They set the risk threshold to match the same overall eligibility as the duration model.
• 9:38
  [SPEAKER_00]: Right, so from a purely statistical view, adding more data should make a model better, right?
• 9:43
  [SPEAKER_00]: I would guess this risk model is the gold standard.
• 9:45
  [SPEAKER_01]: And you'd be right, if you're only looking at the overall population, the risk model improved both sensitivity and specificity for the cohort as a whole.
• 9:53
  [SPEAKER_01]: If your only goal is to maximize statistical efficiency across the country, get the most true positives and the fewest false positives, then yes, this was on paper statistically superior.
• 10:04
  [SPEAKER_00]: But, and I feel a big butt coming, we've already established that statistically superior doesn't automatically mean equitable.
• 10:10
  [SPEAKER_00]: Did this complex model fix the eligibility gaps?
• 10:13
  [SPEAKER_01]: and this is the jaw-dropping finding of the study.
• 10:16
  [SPEAKER_01]: A huge cautionary tale for public health.
• 10:20
  [SPEAKER_01]: Despite being better overall, the RISC-based model was a complete failure on the equity fund.
• 10:26
  [SPEAKER_01]: It actually made the problem worse.
• 10:28
  [SPEAKER_01]: It systematically widened the eligibility gap between Latinos and whites.
• 10:32
  [SPEAKER_00]: It did make it worse.
• 10:33
  [SPEAKER_00]: How bad was it?
• 10:33
  [SPEAKER_01]: It was worse than the system we have now, under the super smart algorithm, only 14.4% of latinos were eligible, compared to 31.3% of whites.
• 10:43
  [SPEAKER_01]: That disparity is even better than the one created by the current, flawed standard.
• 10:47
  [SPEAKER_01]: The algorithm actually amplified the systemic problem for that population.
• 10:51
  [SPEAKER_00]: So the complex model, the one that's supposed to be smarter, was functionally biased against finding cancer in Latinos.
• 10:56
  [SPEAKER_01]: That's what the data shows.
• 10:58
  [SPEAKER_01]: And not only that, the model also had lower sensitivity in Latinos.
• 11:01
  [SPEAKER_01]: It was just worse at finding the actual cancers in that group compared to the simple duration model.
• 11:06
  [SPEAKER_01]: Think about that.
• 11:07
  [SPEAKER_01]: The simple 30 year rule was fundamentally better at identifying lung cancer in Latinos than the complex statistically optimized risk algorithm.
• 11:15
  [SPEAKER_00]: If an algorithm is superior overall, but it systematically fails a specific group of people, doesn't that just make it unusable for public policy?
• 11:25
  [SPEAKER_01]: It raises very, very serious questions.
• 11:27
  [SPEAKER_01]: I mean, it shows how easily historical biases can get baked into these advanced models.
• 11:31
  [SPEAKER_01]: If the data that trains the algorithm or the factors it weighs, don't accurately capture the risk profile of a certain group, the model will just perpetuate inequality, or in this case, make it even worse.
• 11:43
  [SPEAKER_00]: So let's just synthesize all this.
• 11:44
  [SPEAKER_00]: We have the current standard, which creates clear inequities.
• 11:48
  [SPEAKER_00]: We have the sophisticated risk model, which is efficient overall, but makes those inequities even worse.
• 11:53
  [SPEAKER_01]: And then you have the simple, surprisingly effective 30-year duration criteria.
• 11:58
  [SPEAKER_01]: And the study concludes that this model is a very, very strong contender to replace the current standard, because it works on both fronts.
• 12:04
  [SPEAKER_01]: It successfully reduces the eligibility gaps for African-Americans and Latinos, while at the same time it improves sensitivity.
• 12:11
  [SPEAKER_01]: It finds more cancer across all racial groups.
• 12:14
  [SPEAKER_01]: It's just a classic example of how, sometimes a simpler, more direct metric can provide
• 12:22
  [SPEAKER_00]: a really powerful lesson there, that the simplest change could potentially undo decades of structural inequity that was just embedded in a calculation.
• 12:33
  [SPEAKER_01]: And it gives policymakers a clear path forward.
• 12:36
  [SPEAKER_01]: This is easy for doctors to use, it doesn't require some complex calculator, but we should just briefly mention the one limitation the research is like.
• 12:44
  [SPEAKER_00]: Right, the issue of overdiagnosis.
• 12:46
  [SPEAKER_01]: That's right.
• 12:46
  [SPEAKER_01]: The duration criteria found more cancer, which is great, but the study didn't measure over diagnosis, which is finding those slow-growing cancers that might never have actually caused harm.
• 12:56
  [SPEAKER_01]: That's an important cost to consider for any screening program.
• 12:59
  [SPEAKER_01]: But it doesn't really change the core equity findings.
• 13:02
  [SPEAKER_00]: Absolutely not.
• 13:03
  [SPEAKER_00]: The takeaway here is huge.
• 13:05
  [SPEAKER_00]: The criteria we use to decide who gets screamed has this enormous real-world impact on who gets access to life-saving care.
• 13:12
  [SPEAKER_00]: And this study really suggests that a simpler metric, how long you smoked, not how heavily, could be the key to fairer, more effective screening for everyone.
• 13:20
  [SPEAKER_01]: We've seen how these different metrics lead to drastically different outcomes for equity.
• 13:24
  [SPEAKER_01]: So if we connect this to the bigger picture, it raises a really important question for you to think about.
• 13:29
  [SPEAKER_01]: When we're designing these massive public health policies, should the main goal be overall statistical efficiency, like we saw with the risk model, or should maximizing equitable access to care, like the duration model did, be the primary objective, and that is something for you to mull over on your own.
• 13:45
  [SPEAKER_00]: A heavy question, but in necessary one.
• 13:48
  [SPEAKER_00]: Thanks for joining us for this deep dive.