Abstract
Background
There is little guidance on how to personalize recommendations for lung cancer screening that accounts for the variation in expected net benefit from screening. We sought to explore the individual and population implications of identifying net benefit thresholds where lung cancer screening could be encouraged, discouraged, or offered as an option through neutral shared decision making due to screening being highly preference sensitive.
Methods
With a simulated US population of 40- to 80-year-old individuals who had ever smoked, we used microsimulation to estimate individualized quality-adjusted life-years saved with lung cancer screening. We then identified 2 net benefit thresholds for lung cancer screening that account for a range of patient preferences and scientific uncertainties and compared this approach with the current United States Preventive Services Task Force (USPSTF) guidelines.
Results
Our simulated population included 59 million people. In total, 15 million were USPSTF eligible; of those, 53% (8 million) maintained net benefit, even after accounting for unfavorable preferences about screening. Of the USPSTF population, 3% (450 000) are considered low net benefit (routinely discourage) and 47% (7 million) are in an intermediate gray area, where net benefit depends on patient preferences about lung cancer screening (offer-neutral shared decision making). Among adults who had ever smoked, 2.5 million are high net benefit but excluded by current USPSTF criteria; 20.5 million US adults who had ever smoked are intermediate net benefit but currently excluded by USPSTF criteria.
Conclusions
We estimate that half of the USPSTF lung cancer screening–eligible population is in a high-net-benefit group where lung cancer screening could be routinely encouraged. Current lung cancer screening eligibility criteria likely exclude many ever-smokers who are high net benefit and many more with intermediate net benefit.
Introduction
Based on 2 large, randomized trials,1,2 the United States Preventive Services Task Force (USPSTF) concluded that annual lung cancer screening with low-dose computed tomography (CT) has, on average, moderate net benefit for persons aged 50-80 years with a 20 or more pack-year history and who still smoke or had quit within the past 15 years.3 Such an eligibility threshold makes clinical implementation appear simple, but it ignores a key reality of clinical medicine: average benefit is a poor estimate for many patients.4 In fact, all patients who smoke exist along a spectrum of net benefit from lung cancer screening that runs from low to high net benefit, leaving many in an intermediate gray area where the decision to undergo lung cancer screening is not clear-cut and a patient’s informed preferences will determine whether screening is desirable or undesirable (ie, a “preference-sensitive” zone) (Figure 1).5 For individuals faced with a preference-sensitive decision, experts advocate for a neutral approach to shared decision making.6 Currently, the USPSTF universally recommends and Medicare requires shared decision making for all patients before lung cancer screening.7
Figure 1.
Conventional vs tailored guideline recommendations across the spectrum of individualized net benefit. (A) Conventional approaches to guideline recommendations typically use simple criteria, often based on the average net benefit of a given population and traditionally defined using the inclusion criteria of a randomized controlled trial. This approach produces a recommendation that appears dichotomous in nature. (B) Owing to the continuous nature of individualized risk, all patients exist on a spectrum of net benefit. Using a multivariable risk prediction tool that accounts for individualized lung cancer risk and life expectancy enables us to better understand the distribution of net benefit in a population and can enable the establishment of 2 thresholds, thus identifying a gray zone where net benefit is sensitive to a range of patient preferences and scientific uncertainties.
A broad recommendation for a neutral approach to shared decision making is, however, most appropriate when the target population is homogeneous in its estimated net benefit. Spending precious clinical time on detailed shared decision making is of highest priority when the expected net benefit is preference sensitive (Supplementary Methods).6,8 Not all interventions in medicine are highly preference sensitive, and even for a single intervention, the decision may be preference sensitive for some but not others.9,10 It is well established that among patients eligible for lung cancer screening, estimated net benefit varies up to 10-fold due to substantial variation in lung cancer risk and life expectancy.11 Thus, some eligible patients will have a high likelihood of net benefit, and a neutral discussion about lung cancer screening would be less appropriate for such high-net-benefit persons.
Clinicians and patients must ultimately decide whether to be screened. So, how might we help clinicians deal with the continuum of estimated lung cancer risk and life expectancy, to provide more individualized recommendations? For guidance to be useful, thresholds are needed, but a single threshold seems inadequate in a highly diverse population. In this article, we use an established microsimulation model to describe a conceptual framework and systematic approach to providing such guidance, identifying 2 thresholds to define 3 zones: an encourage (recommend) zone; a discourage (recommend against) zone; and an intermediate (preference-sensitive) gray zone, where the informed preferences of a patient are needed to know which course of action is best (Figure 1).
Methods
Study overview and model description
We used a microsimulation model for lung cancer screening to assess the distribution of individualized net benefit from 3 rounds of annual screening. Microsimulation is a type of decision analysis technique that models individual clinical pathways and can better capture the variation in patient characteristics and risk at baseline.12 This particular microsimulation model has been used and described in detail elsewhere.11,13 Briefly, it has a natural history component to simulate lung cancer outcomes in screening’s absence and a screening component based on the National Lung Screening Trial2 and a recent systematic review.14 Updates to the model from prior publications and rationale for certain modelling decisions are in the Supplementary Methods.
Because we were interested in identifying an intermediate gray zone where differences in net benefit and net harm are sensitive to patient preferences, we conducted a sensitivity analysis as our primary analysis by varying assumptions for the amount of disutility (or degree of dislike) assigned to lung cancer screening and its downstream consequences. This process involved assessing outcomes for 2 different scenarios: (1) a favorable disutility scenario representing patients with a low degree of dislike for screening and its downsides and (2) an unfavorable disutility scenario representing patients with a high degree of dislike. Other parameters of the model were left at their base case assumption from prior studies.11 The time horizon was lifetime, and the primary outcome was quality-adjusted life-years (QALYs). A discount rate of 3% was applied, and all analyses were done using R statistical software (R Foundation for Statistical Computing). This study was determined to be exempt by the VA Ann Arbor Healthcare System Institutional Review Board.
Study population and risk stratification
The study population was a nationally representative simulated cohort of adults who had ever smoked aged 40-80 years using data from the National Health Interview Survey. We chose to examine adults who had ever smoked in the United States to understand the variation in lung cancer risk and benefit from lung cancer screening in a broader but still at-risk population and compare this information with the USPSTF population. The National Health Interview Survey is an annual cross-sectional survey that collects self-reported data from noninstitutionalized adults. Sampling weights and covariance among variables were maintained when scaling to the US population.
All patients in the population were stratified by a well-validated, life-gain prediction model for lung cancer screening called Life-Years From Screening–CT (LYFS-CT).15 We chose this prediction model for lung cancer screening in our analysis because it includes prediction of lung cancer death risk as well as life expectancy for smoking adults, both of which are key determinants of potential life-year gains from lung cancer screening.15 Other prediction models were also examined (Table S1).
Incorporating patient preference to define thresholds
Disutility estimates were assigned based on expert opinion of the modeling team informed by a systematic review of values and subsequent lung cancer screening models (Table S2).16 The values in the review were estimated by various methods and included expert opinions of the authors of the studies included in the review. For some values, there were no corresponding values in the literature, and the expert opinion of the current modeling team was used to assign reasonable values based on available information. For example, no utility values reported for chest CT were found in the literature. Although disutility estimates derived from informed patients are ideal, available studies in lung cancer screening have not systematically captured all the disutility information for the model and do not routinely report the distribution of the estimates.
We defined the encourage threshold as occurring where, even despite highly unfavorable preferences toward screening, there is still net benefit or positive QALYs. This finding aligns with conceptual definitions of when a decision is no longer preference sensitive.8 Next, we defined the discourage threshold as occurring where, even despite favorable preferences toward screening overall and only a modest disutility for the screening test itself, there is net harm or negative QALYs. The rationale for this approach is that if a patient cannot be expected to derive benefit from lung cancer screening after accounting for a modest disutility of the screening test itself, then those patients ought to be discouraged from screening.
Specifically, these thresholds were identified by locating where a locally estimated scatterplot smoothing curve fitted to the model outputs crossed zero on a graph plotting life-years gained against QALY gains for each disutility scenario described above. We then examined the data points for this intersection for consistency above or below zero to identify the encourage and discourage thresholds, respectively.
Target populations and outcomes
After encourage and discourage thresholds had been defined, we explored the implications of using these thresholds by identifying 5 key groups: (1) patients eligible by USPSTF criteria who were also high net benefit (above the encourage threshold), (2) patients who were high net benefit but missed by current USPSTF eligibility criteria, (3) patients who were low net benefit (below the discourage threshold) but included by USPSTF, (4) patients who were intermediate net benefit (gray area between the discourage and encourage thresholds) but missed by current USPSTF criteria, and (5) patients who were USPSTF eligible and intermediate net benefit. We chose to highlight race and ethnicity in our results to examine the proportion of these groups included and excluded by the current status quo. Inclusion of race and ethnicity has been shown to be important for lung cancer screening considerations.17
To show the consequences of screening with these decision thresholds, we examined key outcomes and compared them with screening decisions using the 2021 USPSTF lung cancer screening criteria. Key outcomes included lung cancer deaths prevented per 1000 persons, life-years gained per 1000 persons, days of life gained per person screen, and QALYs gained per 1000 persons. Model parameters were set to the base case for these estimates.
Results
Characteristics of the simulated population are reported in Table S3. Figure 2 shows the distribution of predicted life-days gained within the USPSTF-eligible population and the entire study population of adults who had ever smoked aged 40-80 years.
Figure 2.
Distribution of life-days gained within the USPSTF-eligible population and the 40- to 80 year-old ever-smoking population. (A) Distribution of risk profiles across the entire USPSTF-eligible population. The “≥30” bar on the far right of the x-axis represents all persons at or above 30 life-days gained (93th percentile, about 6% of the USPSTF-eligible population). (B) Distribution of risk profiles across the entire population of 40- to 80-year-old adults who had ever smoked. Colored bars represent patients within the discourage zone, preference-sensitive zone, and encourage zone. The “≥30” bar on the far right of the x-axis represents all persons at or above 30 life-days gained (93th percentile, about 7% of the entire study population). USPSTF = US Preventive Services Task Force.
Under current USPSTF guidelines, 15 165 433 persons in the study population of adults who had ever smoked and were 40 to 80 years of age would be eligible for lung cancer screening (25.7% of the study population). If everyone in this group were to undergo lung cancer screening, we estimate 5.8 lung cancer deaths per 1000 screened persons would be prevented, 93.3 life-years per 1000 persons would be gained, 21.6 days of life would be gained per person screened, and 27.0 QALYs would be gained per 1000 persons screened.
Encourage threshold
We identified the encourage threshold at 16.28 life-days gained, which represents the 85th percentile of risk in the study population (Figure 3). Overall, 10 383 194 persons in the study population (17.6%) were in the encourage zone. In Tables S4 and S5, we present results using an alternative, lower disutility for the screening test (‒2 days for every year screened) to define the encourage threshold.
Figure 3.
Establishing encourage and discourage thresholds among 40- to 80-year-old US adults who have ever smoked. Sussman plot demonstrates locally estimated scatterplot smoothing curves and data points of the favorable preferences (low disutility) scenario and unfavorable preferences (high disutility) scenario. Vertical dashed lines represent discourage (left) and encourage (right) thresholds where the associated curves cross zero. Between vertical dashed lines lies the intermediate, preference-sensitive zone; left of the discourage threshold is the discourage zone; and to the right of the encourage threshold is the encourage zone. CT = computed tomography; LYFS-CT = Life-Years From Screening–Computed Tomography; QALY = quality-adjusted life-year.
*The LYFS-CT calculates life-years gained from lung cancer screening using the difference between estimated life expectancy without screening and estimated life expectancy with 3 annual CT screens and 5 years of follow-up. This model incorporates a validated, individualized, all-cause mortality model, the Lung Cancer Death Risk Assessment Tool, and relative risk reduction estimates from the National Lung Screening Trial.15
Included by USPSTF and above the encourage threshold
Among individuals eligible for lung cancer screening by USPSTF guidelines, 52% (about 8 million) were identified above the encourage threshold (Table 1). Compared with the population included by USPSTF but below this threshold, 75% of this high-net-benefit population is aged 60 years or older (vs 39%) and 77% have smoked for more than 40 years (vs 33%) (Table 1). If everyone in this group were to undergo lung cancer screening, we estimate 9 lung cancer deaths per 1000 persons screened would be prevented, 132 life-years per 1000 persons would be gained, 33 days of life would be gained per person screened, and 38 QALYs would be gained per 1000 persons screened (Table 2).
Table 1.
Characteristics of overlapping and nonoverlapping populations using simple (USPSTF) vs multivariable criteria.
| Above encourage, not USPSTF | Below encourage, meet USPSTF | Above encourage, meet USPSTF | Below discourage, meet USPSTF | Below discourage, not USPSTF | Intermediate, meet USPSTF | Intermediate, not USPSTF | |
|---|---|---|---|---|---|---|---|
| Total, No. | 2 528 786 | 7 311 026 | 7 854 408 | 430 035 | 21 673 615 | 6 880 990 | 20 437 420 |
| Race and ethnicity, % | |||||||
| American Indian or Alaska Native | 0.9 | 1.3 | 0.6 | 3.4 | 1.3 | 1.2 | 0.8 |
| Asian or Pacific Islander | 1.9 | 3.2 | 1.2 | 1.2 | 4.5 | 3.3 | 2.5 |
| Hispanic | 3.5 | 9.0 | 2.5 | 17.1 | 14.6 | 8.5 | 7.8 |
| Non-Hispanic Black | 24.9 | 5.6 | 9.7 | 4.4 | 6.2 | 5.7 | 12.9 |
| Non-Hispanic White | 68.8 | 81.0 | 86.0 | 73.8 | 73.3 | 81.4 | 76.0 |
| Sex, % | |||||||
| Male | 53.2 | 57.1 | 57.5 | 60.1 | 52.3 | 56.9 | 53.7 |
| Female | 46.8 | 42.9 | 42.5 | 39.9 | 47.7 | 43.1 | 46.3 |
| Age group, % | |||||||
| 40-44 y | 0 | 0 | 0 | 0 | 29.6 | 0 | 3.2 |
| 45-49 y | 1.9 | 0 | 0 | 0 | 20.5 | 0 | 13.0 |
| 50-54 y | 1.2 | 32.7 | 6.3 | 64.5 | 15.0 | 30.8 | 12.5 |
| 55-59 y | 7.8 | 28.5 | 16.9 | 33.4 | 13.7 | 28.2 | 12.9 |
| 60-64 y | 14.6 | 20.8 | 24.6 | 1.1 | 10.2 | 22.0 | 14.1 |
| 65-69 y | 20.0 | 11.1 | 22.6 | 0.1 | 6.5 | 11.8 | 15.8 |
| 70-74 y | 26.5 | 5.1 | 17.9 | 0.8 | 3.1 | 5.4 | 15.2 |
| 75-79 y | 24.3 | 1.6 | 10.3 | 0.2 | 1.2 | 1.7 | 11.1 |
| Comorbidities, % | |||||||
| 0 | 51.0 | 52.7 | 49.4 | 55.0 | 73.5 | 52.5 | 58.5 |
| 1 | 20.5 | 17.4 | 19.5 | 17.4 | 13.2 | 17.4 | 18.5 |
| 2 | 12.5 | 11.2 | 13.2 | 7.7 | 6.6 | 11.5 | 10.5 |
| 3 | 7.7 | 7.6 | 8.2 | 8.2 | 3.2 | 7.6 | 6.3 |
| ≥4 | 8.3 | 11.1 | 9.6 | 11.6 | 3.6 | 11.1 | 6.3 |
| Lung Cancer Death Risk Assessment Tool decile, % | |||||||
| 1 | 0 | 0.1 | 0 | 1.1 | 27.4 | 0 | 0 |
| 2 | 0 | 1.3 | 0 | 21.7 | 27.4 | 0 | 0 |
| 3 | 0 | 2.9 | 0 | 49.1 | 26.7 | 0 | 0.8 |
| 4 | 0 | 4.7 | 0 | 22.4 | 13.6 | 3.6 | 13.4 |
| 5 | 0 | 8.8 | 0 | 4.5 | 3.7 | 9.1 | 22.0 |
| 6 | 0 | 15.5 | 0 | 0.3 | 1.0 | 16.5 | 22.5 |
| 7 | 0.2 | 26.6 | 0 | 0.9 | 0.2 | 28.2 | 19.3 |
| 8 | 8.9 | 28.3 | 8.4 | 0 | 0.04 | 30.1 | 14.5 |
| 9 | 43.4 | 10.0 | 34.9 | 0 | 0 | 10.6 | 6.7 |
| 10 | 47.5 | 1.8 | 56.7 | 0 | 0 | 1.9 | 0.7 |
| Smoking duration, % | |||||||
| 0-20 y | 1.9 | 1.5 | 0.07 | 9.6 | 67.7 | 1.0 | 26.4 |
| 21-30 y | 11.8 | 17.4 | 1.4 | 77.0 | 28.4 | 13.7 | 31.3 |
| 31-40 y | 29.2 | 48.3 | 17.8 | 10.9 | 3.8 | 50.6 | 33.8 |
| 41-50 y | 32.6 | 29.6 | 51.6 | 1.9 | 0.06 | 31.4 | 7.5 |
| 51-60 y | 20.9 | 3.1 | 25.0 | 0.7 | 0.01 | 3.3 | 0.9 |
| Pack-year history, % | |||||||
| 0-9 | 29.0 | 0 | 0 | 0 | 62.1 | 0 | 36.0 |
| 10-19 | 35.7 | 0 | 0 | 0 | 23.7 | 0 | 34.2 |
| 20-29 | 2.6 | 43.0 | 12.9 | 90.0 | 12.1 | 40.1 | 13.0 |
| 30-39 | 7.1 | 33.8 | 17.1 | 6.0 | 1.3 | 35.1 | 9.8 |
| 40-49 | 6.5 | 11.5 | 27.5 | 1.7 | 0.5 | 12.1 | 3.1 |
| 50-59 | 5.5 | 3.6 | 16.2 | 1.9 | 0.2 | 3.8 | 1.5 |
| ≥60 | 13.5 | 8.6 | 26.2 | 0.4 | 0.2 | 9.1 | 2.3 |
| Years since quitting, % | |||||||
| 0-5 | 65.7 | 64.0 | 86.5 | 18.6 | 28.8 | 66.8 | 43.3 |
| 6-10 | 1.3 | 20.1 | 8.3 | 33.0 | 10.3 | 19.3 | 4.5 |
| 11-15 | 0.3 | 16.0 | 5.2 | 48.4 | 9.4 | 13.9 | 3.6 |
| 16-20 | 15.0 | 0 | 0 | 0 | 12.0 | 0 | 10.3 |
| 21-25 | 6.8 | 0 | 0 | 0 | 8.0 | 0 | 7.2 |
| 26-30 | 7.3 | 0 | 0 | 0 | 10.6 | 0 | 10.2 |
| >30 | 3.7 | 0 | 0 | 0 | 20.8 | 0 | 21.0 |
Abbreviation: USPSTF = US Preventive Services Task Force.
Table 2.
Population outcomes of overlapping and nonoverlapping groups.
| Above encourage, not USPSTF | Below encourage, meet USPSTF | Above encourage, meet USPSTF | Below discourage, meet USPSTF | Below discourage, not USPSTF | Intermediate, meet USPSTF | Intermediate, not USPSTF | |
|---|---|---|---|---|---|---|---|
| Total, No. | 2 528 786 | 7 311 026 | 7 854 408 | 430 035 | 21 673 615 | 6 880 990 | 20 437 420 |
| Lung cancer deaths prevented per 1000 persons screened | 5.9 | 2.2 | 9.1 | 0.6 | 0.3 | 2.3 | 1.4 |
| Life-years gained per 1000 persons screened | 83.9 | 51.9 | 131.8 | 18.2 | 12.9 | 54.0 | 37.1 |
| Days of life gained per person screeneda | 23.2 | 9.6 | 32.8 | 2.2 | 1.7 | 10.1 | 6.9 |
| QALYs gained per 1000 persons screened | 21.9 | 15.0 | 38.2 | 2.8 | 0.7 | 15.7 | 9.6 |
Abbreviations: QALY = quality-adjusted life year; USPSTF = US Preventive Services Task Force.
Estimate obtained from the R lcrisk package, not the result of the microsimulation model; thus, the relationship between this estimate and the rest of the rows in the table is not scalar.
Above encourage threshold but excluded by USPSTF
We estimate 2.5 million persons (4.3% of the study population) in the United States are above the encourage threshold but excluded by current USPSTF criteria. When we assessed the characteristics of this population, we found that 25% were non-Hispanic Black compared with 7.7% of the current USPSTF-eligible population under the encourage threshold (Table 1). Adding this population to the current non-Hispanic Black population eligible by USPSTF criteria would result in a 54% increase in the size of the lung cancer screening–eligible Black population. Among the high-net-benefit population currently excluded by USPSTF criteria, about 50 000 (1.9%) are aged 40-49 years, and 32.8% quit smoking more than 15 years ago; 85% were 60 years of age or older compared with 38.6% for persons USPTSF eligible but below the encourage threshold (Table 1).
If everyone in this group were to undergo lung cancer screening, about 6 lung cancer deaths per 1000 persons screened would be prevented, 84 life-years per 1000 persons would be gained, 23 days of life would be gained per person screened, and 22 QALYs would be gained per 1000 persons screened (Table 2).
Discourage threshold
We identified the discourage threshold at 3.28 life-days gained, representing the 33rd percentile of risk in the study population (Figure 3). More than 20 million persons in the study population (36.5%) are in the discourage zone. Table 2 shows the population outcomes of screening in this group.
Below discourage threshold, included by USPSTF
Of the individuals in the discourage zone, 2.8% (430 035 people) are USPSTF eligible; 98% of the persons in this group were aged 50-59 years, and almost one-quarter were in the first and second deciles of risk (Table 1). If everyone in this group were to undergo lung cancer screening, fewer than 1 lung cancer death per 1000 persons screened would be prevented, about 18 life-years per 1000 persons would be gained, 2 days of life would be gained per person screened, and 3 QALYs would be gained per 1000 persons screened (Table 2).
Intermediate, preference-sensitive zone
In all, 27 318 411 people in the study population were in the intermediate, preference-sensitive gray zone. Almost 7 million of those persons in the preference-sensitive zone were USPSTF eligible, which is 44.8% of the current USPSTF-eligible population.
Intermediate, preference-sensitive zone but excluded by USPSTF
About 20.5 million adults who had ever smoked in the preference-sensitive zone are currently excluded by USPSTF guidelines, which comprises 35% of the entire study population (Table 2). About 13% of that population is non-Hispanic Black compared with the 7.7% in the current USPSTF-eligible population, which, if added to the current USPSTF-eligible population, would more than double the non-Hispanic Black population (225% increase). Sixteen percent of this population is aged 40-49 years, 56% are 60 years of age or older, and approximately 50% had quit smoking more than 15 years previously. Outcomes for this group are presented in Table 2.
In Tables S6 and S7, we also present the population characteristics and outcomes for the entire population above the lower discourage threshold, which includes both intermediate (preference-sensitive) gray zone and encourage zone individuals.
Discussion
We present a conceptual framework using microsimulation to identify an intermediate gray zone for lung cancer screening by setting 2 thresholds: 1 below which screening is discouraged and another above which screening is encouraged. We estimated that 52% of the current USPSTF-eligible population can expect net benefit from lung cancer screening, even after accounting for unfavorable preferences about lung cancer screening. Despite present recommendations to conduct shared decision making for all individuals eligible for lung cancer screening, such screening is less preference sensitive for this high-net-benefit group; thus, neutral shared decision making may be less appropriate. Based on our analysis, current guidelines do well in excluding many low-risk, low-net-benefit patients, but they also exclude many persons with appreciable lung cancer risk and reasonable life expectancy who have at least moderate net benefit: more than 2 million adults who are high net benefit and more than 20 million adults at intermediate net benefit for whom lung cancer screening may be reasonable if it aligns with their preferences. Disproportionately, individuals in these intermediate- or high-net-benefit groups who are excluded by current criteria are Black, are older (≥ 60 years), have smoked for a longer duration but with less intensity (<20 pack-years), and quit more than 15 years previously.
Traditional approaches to producing guideline recommendations18,19 typically focus on setting a single threshold for a specific population that creates 2 seemingly distinct groups: individuals for whom the intervention should be recommended (weakly or strongly) and individuals for whom it should not. Such simple criteria can quickly communicate to busy clinicians which broad populations are likely to benefit, on average. We found, however, that simple criteria may have considerable limitations for guiding lung cancer screening decisions for individual patients with variable lung cancer risk and life expectancy and that simple criteria likely exclude many individuals who have a substantial chance of benefitting from lung cancer screening and many more for whom lung cancer screening could be reasonable. Setting 2 risk thresholds as opposed to 1 enables us to better address the continuous nature of net benefit from lung cancer screening, account for a range of patient preferences, and identify persons for whom the decision to undergo lung cancer screening is most preference sensitive.
Although individualized risk prediction has the potential to introduce greater complexity into lung cancer screening recommendations and decisions, the 2-threshold approach can still clearly communicate a recommended course of action to clinicians: Patients in the encourage zone could be routinely encouraged to undergo lung cancer screening, and persons in the preference-sensitive zone could undergo a neutral shared decision making process to further elucidate patient preferences and determine the best course of action, with persons in the discourage zone being ineligible.
This approach also provides a path for improving the delivery of interventions to promote lung cancer screening uptake.20 Lung cancer screening rates are lower in the United States than are other US cancer screening programs,21 and persons in the encourage zone could be prioritized for effective but limited interventions known to increase screening, such as screening coordinators and ride programs.22 Related research has shown that an individualized, net benefit–based approach to lung cancer screening can increase uptake and close important care gaps.23-25
We identified large populations of adults who had ever smoked and may benefit from lung cancer screening but are currently excluded by USPSTF criteria. We estimated that persons excluded but above our encourage threshold would have more lung cancer deaths prevented and QALY gains per 1000 persons screened than many who are currently eligible by USPSTF criteria. Further, many more who are excluded by USPSTF criteria but in our intermediate zone had more estimated lung cancer deaths prevented and QALY gains per 1000 persons screened than some persons currently meeting USPSTF eligibility. These striking results suggest that current criteria result in substantial inequities of care and contradict the equal management of equal risk principle.26
These results are similar to those from a recent study that also identified a high-net-benefit group for lung cancer screening using fast-and-frugal trees.27 As described in that article and confirmed here, the differences in populations captured by USPSTF criteria and the prediction model likely have to do with the USPSTF’s use of pack-years as a surrogate for cumulative smoking exposure vs the LYFS-CT model, which weighs duration of tobacco use separately from smoking intensity. Smoking duration seems to be a more considerable risk factor for lung cancer that should be considered separately from smoking intensity.28,29
We estimate that more than 20 million adults who had ever smoked and are between 40 and 80 years of age fall within the intermediate gray zone but are excluded by current criteria. Adding this population to the current USPSTF-eligible population would more than double the current population eligible for lung cancer screening, meaning that there are many more patients than currently eligible who would potentially benefit from being offered lung cancer screening through a neutral shared decision making discussion. Opening eligibility to a broader population of former smokers is consistent with new American Cancer Society guidelines that removed the years since quitting criterion.30
In practice, identifying more people for whom lung cancer screening may be reasonable could present substantial time allocation problems to already busy clinicians. Proposals for more efficient ways to provide shared decision making are available,31 and the delivery of patient decision aids before an appointment would be essential to effective shared decision making. Nonetheless, adding this group would substantially and disproportionately increase the number of non-Hispanic, Black patients eligible for lung cancer screening, which may provide a mechanism for decreasing important lung cancer disparities known to exist among this population.32 Prior studies have shown that the 2021 USPSTF lung cancer screening criteria could potentially reduce racial and ethnic disparities in lung cancer outcomes,33,34 but our analyses align with more recent studies that suggest an even greater potential to reduce these disparities through a prediction-based approach.17,27,35 Further work to improve lung cancer screening uptake and adherence among these high-risk groups is needed.
Our study has several limitations. Setting thresholds is ultimately a value judgment. Simulation and prediction can provide useful information to decision makers, but models have their own uncertainties and imprecision regarding input parameters and risk model calibration error.11 Setting the range of patient disutility estimates for the burden and complications for any type of screening is difficult, but further research to better estimate these disutilities is needed.36-39 We tried to err on the side of a broad range of individual patient disutilities; thus, we may have overestimated the width of the preference-sensitive zone. The thresholds we present should be viewed as rough guideposts.
In conclusion, simple eligibility criteria for lung cancer screening exclude many patients with higher-than-average net benefit from lung cancer screening and exclude even more for whom the decision to undergo lung cancer screening is likely preference sensitive. More than half of currently eligible patients could receive stronger encouragement to undergo lung cancer screening rather than neutral shared decision making. To realize these potential benefits of lung cancer screening, greater resources and better strategies and interventions are needed to improve the uptake of screening among persons at risk.
Supplementary Material
Acknowledgments
The funder did not play a role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the US Department of Veterans Affairs or the US government. Preliminary results of this study were presented at the 2023 Society for Medical Decision Making annual meeting.
Contributor Information
Joshua B Rager, Division of General Internal Medicine and Geriatrics, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, United States; Center for Health Services Research, Regenstrief Institute, Indianapolis, IN, United States; Indiana University Center for Bioethics, Indiana University School of Medicine, Indianapolis, IN, United States.
Pianpian Cao, Department of Public Health, College of Health and Human Sciences, Purdue University, West Lafayette, IN, United States.
Rodney A Hayward, Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI, United States; Division of General Medicine, Department of Medicine, University of Michigan Medical School, Ann Arbor, MI, United States.
Rafael Meza, BC Cancer Research Institute, Vancouver, British Columbia, Canada.
Hormuzd A Katki, Division of Cancer Epidemiology and Genetics, US Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, MD, United States.
Jeremy B Sussman, Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI, United States; Division of General Medicine, Department of Medicine, University of Michigan Medical School, Ann Arbor, MI, United States.
Tanner J Caverly, Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI, United States; Division of General Medicine, Department of Medicine, University of Michigan Medical School, Ann Arbor, MI, United States; Department of Learning Heath Sciences, University of Michigan Medical School, Ann Arbor, MI, United States.
Author contributions
Joshua B. Rager (Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing—original draft), Pianpian Cao (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Writing—review & editing), Rodney A. Hayward (Conceptualization, Investigation, Methodology, Supervision, Writing—review & editing), Rafael Meza (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Supervision, Writing—review & editing), Hormuzd A. Katki (Methodology, Resources, Software, Supervision, Writing—review & editing), Jeremy B. Sussman (Conceptualization, Methodology, Supervision, Writing—review & editing), and Tanner J. Caverly (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Visualization, Writing—original draft, Writing—review & editing)
Supplementary material
Supplementary material is available at JNCI: Journal of the National Cancer Institute online.
Funding
This work was supported by the US Department of Veterans Affairs Health Systems Research (VA HSR IIR 21-152).
Conflicts of interest
T.J.C. reports an open source Apache 2.0 license for a freely available decision support tool (screenlc.com) for which he does not receive compensation. H.A.K., who is a JNCI editorial board member and co-author of this article, was not involved in the editorial review or decision to publish the manuscript.
Data availability
The patient and population data used for this study are available publicly through the National Health Interview Survey at cdc.gov. The microsimulation model is not publicly available; for those interested in this model, please visit cisnet.cancer.gov.
References
- 1. de Koning HJ, van der Aalst CM, de Jong PA, et al. Reduced lung-cancer mortality with volume CT screening in a randomized trial. N Engl J Med. 2020;382:503-513. 10.1056/NEJMoa1911793 [DOI] [PubMed] [Google Scholar]
- 2. National Lung Screening Trial Research Team; Aberle DR, Adams AM, Berg CD, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011;365:395-409. 10.1056/NEJMoa1102873 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Krist AH, Davidson KW, Mangione CM, et al. ; US Preventive Services Task Force. Screening for lung cancer: US preventive services task force recommendation statement. JAMA. 2021;325:962-970. 10.1001/jama.2021.1117 [DOI] [PubMed] [Google Scholar]
- 4. Kent DM, Hayward RA. Limitations of applying summary results of clinical trials to individual patients the need for risk stratification. JAMA. 2007;298:1209-1212. 10.1001/jama.298.10.1209 [DOI] [PubMed] [Google Scholar]
- 5. Quanstrum KH, Hayward RA. Lessons from the mammography wars. N Engl J Med. 2010;363:1076-1079. 10.1056/NEJMsb1002538 [DOI] [PubMed] [Google Scholar]
- 6. Elwyn G, Frosch D, Rollnick S. Dual equipoise shared decision making: definitions for decision and behaviour support interventions. Implement Sci. 2009;4:75. 10.1186/1748-5908-4-75 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Hoffman RM, Reuland DS, Volk RJ. The Centers for Medicare & Medicaid Services requirement for shared decision-making for lung cancer screening. JAMA. 2021;325:933-934. 10.1001/jama.2021.1817 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Whitney SN, McGuire AL, McCullough LB. A typology of shared decision making, informed consent, and simple consent. Ann Intern Med. 2004;140:54-59. 10.7326/0003-4819-140-1-200401060-00012 [DOI] [PubMed] [Google Scholar]
- 9. Elwyn G, Price A, Franco JVA, Gulbrandsen P. The limits of shared decision making. BMJ Evid Based Med. 2023;28:218-221. 10.1136/bmjebm-2022-112089 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. van der Horst DEM, Garvelink MM, Bos WJW, Stiggelbout AM, Pieterse AH. For which decisions is shared decision making considered appropriate?—A systematic review. Patient Educ Couns. 2023;106:3-16. 10.1016/j.pec.2022.09.015 [DOI] [PubMed] [Google Scholar]
- 11. Caverly TJ, Cao P, Hayward RA, Meza R. Identifying patients for whom lung cancer screening is preference-sensitive. Ann Intern Med. 2018;169:1-9. 10.7326/M17-2561 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Siebert U, Alagoz O, Bayoumi AM, et al. ; ISPOR-SMDM Modeling Good Research Practices Task Force. State-transition modeling: a report of the ISPOR-SMDM modeling good research practices task force-3. Value in Health. 2012;15:812-820. 10.1016/j.jval.2012.06.014 [DOI] [PubMed] [Google Scholar]
- 13. BCCRI-LunCan (BCCRI). Cancer Intervention and Surveillance Modeling Network. Accessed August 12, 2025. https://cisnet.cancer.gov/resources/model-registry/lung-models/BCCRI-LungCan.html#/
- 14. Jonas DE, Reuland DS, Reddy SM, et al. Screening for lung cancer with low-dose computed tomography: updated evidence report and systematic review for the US Preventive Services Task Force. JAMA. 2021;325:971-987. 10.1001/jama.2021.0377 [DOI] [PubMed] [Google Scholar]
- 15. Cheung LC, Berg CD, Castle PE, Katki HA, Chaturvedi AK. Life-gained–based versus risk-based selection of smokers for lung cancer screening. Ann Intern Med. 2019;171:623-632. 10.7326/M19-1263 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Earle CC, Chapman RH, Baker CS, et al. Systematic overview of cost-utility assessments in oncology. J Clin Oncol. 2000;18:3302-3317. 10.1200/JCO.2000.18.18.3302 [DOI] [PubMed] [Google Scholar]
- 17. Landy R, Gomez I, Caverly TJ, et al. Methods for using race and ethnicity in prediction models for lung cancer screening eligibility. JAMA Network Open. 2023;6:e2331155-e2331155. 10.1001/jamanetworkopen.2023.31155 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Barry MJ, Wolff TA, Pbert L, et al. Putting evidence into practice: an update on the us preventive services task force methods for developing recommendations for preventive services. Ann Fam Med. 2023;21:165-171. 10.1370/afm.2946 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Guyatt GH, Oxman AD, Vist GE, et al. ; GRADE Working Group. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336:924-926. 10.1136/bmj.39489.470347.AD [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Wiener RS, Gould MK. Selecting candidates for lung cancer screening: implications for effectiveness, efficiency, equity, and implementation. Ann Intern Med. 2023;176:413-414. 10.7326/M23-0230 [DOI] [PubMed] [Google Scholar]
- 21. Maki KG, Tan NQP, Toumazis I, Volk RJ. Prevalence of lung cancer screening among eligible adults in 4 US states in 2021. JAMA Network Open. 2023;6:e2319172. 10.1001/jamanetworkopen.2023.19172 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Sayani A, Ali MA, Dey P, et al. Interventions designed to increase the uptake of lung cancer screening: an equity-oriented scoping review. JTO Clin Res Rep. 2023;4:100469. 10.1016/j.jtocrr.2023.100469 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Caverly TJ, Wiener RS, Kumbier K, Lowery J, Fagerlin A. Prediction-augmented shared decision-making and lung cancer screening uptake. JAMA Network Open. 2024;7:e2419624. 10.1001/jamanetworkopen.2024.19624 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Kukhareva PV, Li H, Caverly TJ, et al. Lung cancer screening before and after a multifaceted electronic health record intervention: a nonrandomized controlled trial. JAMA Network Open. 2024;7:e2415383. 10.1001/jamanetworkopen.2024.15383 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Kukhareva PV, Li H, Caverly TJ, et al. Implementation of lung cancer screening in primary care and pulmonary clinics: pragmatic clinical trial of electronic health record-integrated everyday shared decision-making tool and clinician-facing prompts. Chest. 2023;164:1325-1338. 10.1016/j.chest.2023.04.040 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Perkins RB, Guido RS, Castle PE, et al. ; 2019 ASCCP Risk-Based Management Consensus Guidelines Committee. 2019 ASCCP risk-based management consensus guidelines for abnormal cervical cancer screening tests and cancer precursors. J Low Genit Tract Dis. 2020;24:102-131. 10.1097/LGT.0000000000000525 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Kearney LE, Belancourt P, Katki HA, et al. The development and performance of alternative criteria for lung cancer screening. Ann Intern Med. 2024;177:1222-1232. 10.7326/M23-3250 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Pleasants RA, Rivera MP, Tilley SL, Bhatt SP. Both duration and pack-years of tobacco smoking should be used for clinical practice and research. Ann Am Thorac Soc. 2020;17:804-806. 10.1513/AnnalsATS.202002-133VP [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Remen T, Pintos J, Abrahamowicz M, Siemiatycki J. Risk of lung cancer in relation to various metrics of smoking history: a case-control study in Montreal. BMC Cancer. 2018;18:1275. 10.1186/s12885-018-5144-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Wolf AMD, Oeffinger KC, Shih TYC, et al. Screening for lung cancer: 2023 guideline update from the American Cancer Society. CA Cancer J Clin. 2024;74:50-81. 10.3322/caac.21811 [DOI] [PubMed] [Google Scholar]
- 31. Caverly TJ, Hayward RA. Dealing with the lack of time for detailed shared decision-making in primary care: everyday shared decision-making. J Genit Intern Med. 2020;35:3045-3049. 10.1007/s11606-020-06043-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. CA Cancer J Clin. 2019;69:7-34. 10.3322/caac.21551 [DOI] [PubMed] [Google Scholar]
- 33. Ritzwoller DP, Meza R, Carroll NM, et al. Evaluation of population-level changes associated with the 2021 US Preventive Services Task Force lung cancer screening recommendations in community-based health care systems. JAMA Network Open. 2021;4:e2128176. 10.1001/jamanetworkopen.2021.28176 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Reese TJ, Schlechter CR, Potter LN, et al. Evaluation of revised US Preventive Services Task Force lung cancer screening guideline among women and racial/ethnic minority populations. JAMA Network Open. 2021;4:e2033769. 10.1001/jamanetworkopen.2020.33769 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Choi E, Ding VY, Luo SJ, et al. Risk model–based lung cancer screening and racial and ethnic disparities in the US. JAMA Oncol. 2023;9:1640-1648. 10.1001/jamaoncol.2023.4447 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. van den Bergh KAM, Essink-Bot ML, Borsboom GJJM, et al. Short-term health-related quality of life consequences in a lung cancer CT screening trial (NELSON). Br J Cancer. 2010;102:27-34. 10.1038/sj.bjc.6605459 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. van den Bergh KAM, Essink-Bot ML, Borsboom GJJM, Scholten ET, van Klaveren RJ, de Koning HJ. Long-term effects of lung cancer computed tomography screening on health-related quality of life: the NELSON trial. Eur Respir J. 2011;38:154-161. 10.1183/09031936.00123410 [DOI] [PubMed] [Google Scholar]
- 38. Rasmussen JF, Siersma V, Malmqvist J, Brodersen J. Psychosocial consequences of false positives in the Danish Lung Cancer CT Screening Trial: a nested matched cohort study. BMJ Open. 2020;10:e034682. 10.1136/bmjopen-2019-034682 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Ngo PJ, Cressman S, Behar-Harpaz S, Karikios DJ, Canfell K, Weber MF. Applying utility values in cost-effectiveness analyses of lung cancer screening: A review of methods. Lung Cancer. 2022;166:122-131. 10.1016/j.lungcan.2022.02.009 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The patient and population data used for this study are available publicly through the National Health Interview Survey at cdc.gov. The microsimulation model is not publicly available; for those interested in this model, please visit cisnet.cancer.gov.