Abstract
The National Lung Screening Trial (NLST) demonstrated a 20% reduction in lung cancer mortality for screening with low-dose computed tomography versus chest radiography. The major NLST eligibility criteria were age 55–74, a 30 + pack year smoking history and current smoking status or having quit in the last 15 years. We utilized data from SEER (Surveillance, Epidemiology and End Results), the United States (US) Census and the National Health Interview Survey, as well as two statistical models of lung cancer risk, to estimate the proportion of the total US population and of those currently diagnosed with lung cancer that would be covered by the NLST and other suggested eligibility criteria. For the NLST criteria, 26.7% of lung cancers and 6.2% of the population (over 40) were covered. A criterion of ever smokers aged 50–79 would cover 68% of the cancers while screening 30% of the (over 40) population. To extend recommended screening beyond the NLST eligibility criteria, two questions are key. First, can the 20% mortality reduction observed in NLST be extrapolated to populations at moderately lower risk? Second, given that such an extrapolation is valid, what background incidence rate is high enough for the balance between the benefits and harms of screening to be favourable? Further research on these questions is needed.
INTRODUCTION
Recently, the results of the National Lung Screening Trial (NLST) were published, showing a 20% reduction in lung cancer mortality for three rounds of annual screening with low-dose computed tomography (LDCT) compared with chest radiography. 1 The major eligibility criteria for NLST were age 55–74, a 30 + pack year smoking history and current smoking status or having quit in the last 15 years. The above then comprise the default criteria for recommending routine LDCT screening in the United States (US) population, although one organization has recommended screening to a slightly wider group. 2,3 We sought to determine what proportion of all those persons currently diagnosed with lung cancer in the US is covered under these and other suggested criteria.
METHODS
We utilized 2007–2008 incidence rates from SEER 17 (Surveillance, Epidemiology and End Results) and 2010 US Census population counts, to estimate the total number of annual US lung cancer cases by five year age groups. 4,5 We used the 2010 National Health Interview Survey (NHIS) to estimate the proportion of the population (by age group) in various smoking history categories, including that defined by the NLST eligibility criteria (i.e. 30 + pack years and either current smoking or quitting within the past 15 years). The NHIS captures respondents’ past and present cigarette smoking as well as demographic and behavioural data. 6 Finally, we utilized statistical models of lung cancer risk as a function of smoking history and other covariates (e.g. age, sex, education) to compute the relative risk (RR) of lung cancer, in each age category, for the various smoking history categories (with referent group never smokers). The RRs from two models, both of which were fitted to populations that included never smokers in addition to current and former smokers, were averaged to give the RRs for the current analysis. 7,8
The proportion of all lung cancers, in an age group, in the ith smoking history category was then calculated as fi RR i /[∑ j fj RR j ] where fi and RR i equal the fraction in the ith category and the RR in the ith category, respectively. These proportions were multiplied by the total number of incident lung cancers in the age group and summed over age groups as appropriate. The lung cancer incidence rate in the ith smoking history category (in an age group) was calculated as Ii = bRR i where b = I/[∑ j fj RR j ] and I is the SEER (age-specific) incidence rate. The lung cancer incidence rate for concatenations of age by smoking history categories was calculated as a weighted average of the category-specific rates, with weights proportional to population size.
RESULTS
Proportion of lung cancers, proportion of population screened (aged over 40) and lung cancer incidence rate by smoking and age category
PY, per year.
*NLST eligibility criteria (along with age 55–74)
Bold values indicate the NLST eligibility criteria in terms of smoking and age
Categories B, C and D are not mutually exclusive. B and C each include category A, and D includes A, B and C. Values for the subsets B–A, etc. (e.g. 30 + pack years and not NLST eligible), are derived by subtraction for % of cancers and % of population screened; incidence rate is derived as R B–A = (R B P B − R A P A)/(P B − P A) where R is incidence rate and P is % of population in subscripted category.
DISCUSSION
Only 26.7% of subjects currently being diagnosed with lung cancer meet the strict NLST eligibility criteria. Although screening in this group would be efficient, with 26.7% of the lung cancers covered while only screening 6% of the population (aged 40+), the overall benefit of the screening would be modest, although still significant. With a 20% mortality reduction, and, say, 80% compliance, the overall lung cancer mortality reduction would be on the order of 5%. Expanding to age 50–79 and all persons with a 30 + pack year history would increase the lung cancer coverage to 46.3% while increasing the population percentage screened to 12.1%.
There are two key questions involved with attempting to extend the recommended screening population beyond those meeting the NLST eligibility criteria. First, can the lung cancer mortality RR observed in NLST be extrapolated to populations at moderately lower risk, in terms of age and smoking history? Although in a strict sense, only a randomized controlled trial performed with the given expanded eligibility criteria could provide definitive proof, most would agree that such an extrapolation is reasonable for modest expansions of the age criterion (including those 50–54 and/or 75–79) and smoking history criterion (category B and C from the Table).
Second, given that the above extrapolation is valid, is the background incidence rate in that population high enough, assuming the same percent mortality reduction as in NLST, such that the balance between the benefits and harms of screening are favorable? For a given percent mortality reduction, the number needed to screen to prevent a lung cancer death (NNS) is directly proportional to the inverse of the background (i.e. in the absence of screening) lung cancer mortality rate in the relevant population. Because lung cancer has such a high case fatality rate, the mortality rate will be approximately the same (slightly lower) as the incidence rate. Therefore, again assuming a constant percent mortality benefit, the ratio of NNS for the categories considered here will be approximately the inverse of the corresponding incidence rate ratio. For age 55–74 and current smokers or those quitting within the past 15 years regardless of pack years (category C), the NNS would be 1.33 times greater than the NNS for the NLST eligibility criteria (i.e. 579/436). Considering only the subset of the above group (age 55–74 and category C) that did not meet the NLST criteria, i.e. those with under 30 pack years, the NNS would be 2.2 times greater, based on an incidence rate (per 100,000) of 264. For context, the NNS in NLST was 320. 1 If there were a cut-off NNS level above which screening were no longer favourable in terms of benefit versus harms, then a low enough background mortality rate, even with the same percent mortality reduction, could lead to a decision to not recommend screening. Note that NNS measures lives saved but not life-years saved, so when comparing the NNS across different age groups, a similar NNS may represent more life-years saved in a younger population.
Diagnostic follow up from false-positive screens, excess radiation and over-diagnosed cancers all contribute to the medical harms of lung cancer screening with LDCT. The (per-capita) financial costs of screening, including those downstream from the screen itself, are also considerable. False-positive rates in NLST were about 26% in the first two screening rounds and 16% in the last round; cancer detection rates averaged slightly under 1% per round. 1 Currently, a cost–benefit analysis of LDCT screening is being conducted by NLST researchers; results from this will further inform the tradeoffs of screening in different risk groups.
CONCLUSION
There are many factors to consider as individuals, health-care providers, policy-makers and insurers weigh implementing LDCT screening for lung cancer. Considerations of the balance of the population needed to be screened, along with the harms and costs, against the potential expected reduction in the population mortality burden from lung cancer, the current leading cause of cancer death, are useful.