Chest ACCP Member Benefits
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Guest Access | Sign In via User Name/Password
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (24)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chirikos, T. N.
Right arrow Articles by Clark, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chirikos, T. N.
Right arrow Articles by Clark, R.
(Chest. 2002;121:1507-1514.)
© 2002 American College of Chest Physicians

Screening for Lung Cancer With CT*

A Preliminary Cost-effectiveness Analysis

Thomas N. Chirikos, PhD; Todd Hazelton, MD; Melvin Tockman, MD, PhD and Robert Clark, MD

* From the H. Lee Moffitt Cancer Center and Research Institute at the University of South Florida, Tampa, FL.

Correspondence to: Robert Clark, MD, H. Lee Moffitt Cancer Center and Research Institute at the University of South Florida, 12902 Magnolia Dr, Tampa, FL 33612; e-mail: clark{at}moffitt.usf.edu


   Abstract
TOP
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Study objective: To appraise the potential cost-effectiveness of lung cancer screening with CT.

Study design: Incremental cost-effectiveness ratios are estimated for two hypothetical cohorts followed up over time. One cohort was screened over the first 5 years of the study period; the other cohort received usual care. Cost streams are projected for each cohort under alternative sets of parameters/ assumptions and from the perspective of national payer groups. Cohort cost differentials arise as a result of screening and variations in stage-specific treatment. Cohort life expectancies are also projected, and they too differ as a consequence of variations in the stage distribution at diagnosis. The ratios of these cost and life-expectancy differences are used to judge the expected economic value of screening.

Results: Results are analyzed for a "worst-case" scenario, ie, with the highest cost and lowest yield assumptions. Under these conditions, screening with CT costs approximately $48,000 per life-year gained, if screening results in 50% of lung cancers detected at localized stage. Smaller proportions of cancer detected at a localized stage result in higher cost-effectiveness ratios, and vice versa.

Conclusion: If screening for lung cancer is effective, it is likely to be cost-effective if the screening process can detect > 50% of cancers at localized stage.

Key Words: cost-effectiveness • CT • lung cancer • screening protocol


   Introduction
TOP
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Lung cancer is the leading cause of cancer deaths for both men and women in the United States. This year, 154,900 men and women will die from lung cancer, more deaths than from breast, prostate, cervix, and colon cancers combined.1 It is not a coincidence that these other four cancers have screening programs in place, permitting early detection and better survival.

In spite of advances in treatment, the survival rate for lung cancer has not appreciably improved in the past 40 years. In the United States, the overall 5-year survival rate remains a dismal 13%, and the 10-year survival rate is 7%.2 Survival is related to stage at presentation. The 5-year survival rate for patients with localized (node-negative) tumors is > 50%,2 3 4 while the survival rate for T1N0M0 disease is 67 to 80%.5 6

Unfortunately, < 15% of lung cancers in the United States are detected early. The current stage distribution at clinical presentation is heavily weighted toward advanced disease, with only 12% of patients

presenting with stage I (T1N0) lesions, and 15% with stage II cancers (T1N1, T2N0, or T2N1).3 Effective screening strategies, therefore, have potential for significant beneficial impact on lung cancer survival and mortality rates.

At this time, no organization recommends routine screening for lung cancer, either among the general population or in individuals who are at higher risk due to tobacco or occupational exposures.7 Primary prevention (ie, prevention and cessation of cigarette smoking) is the only cancer control strategy currently recommended. However, new cases of lung cancer are now as common in former smokers as in current smokers.8 Secondary prevention with screening would be desirable for the millions of smokers and former smokers who remain at high risk for lung cancer.

Four randomized, controlled trials of lung cancer screening with chest radiography alone or in combination with sputum cytology demonstrated no mortality reduction benefit to screened groups.9 10 11 12 13 14 15 16 17 18 However, these trials had several limitations. For one thing, they were limited by design, statistical power, and screening contamination of the control groups. For another, they enrolled only men, although currently 47% of new lung cancer cases and 43% of lung cancer deaths occur in women. For yet another, when these trials were conducted, squamous cell carcinoma was the predominant histology of lung cancer, occurring in the central bronchi, and less easily detected by chest radiography. In the past decade, this pattern has changed, for unknown reasons, to one in which adenocarcinoma is the predominant histology. Adenocarcinoma is typically more peripherally located and more easily detectable by lung imaging techniques. Therefore, it is reasonable to consider new imaging strategies for lung cancer screening.

A promising new approach using low-dose, single-breath, helical CT has again raised interest in lung cancer screening.19 20 Henschke et al,19 in their Early Lung Cancer Action Project (ELCAP), screened 1,000 smokers and former smokers >= 60 years old with CT and detected 27 lung cancers, > 80% of which were stage IA tumors. Although CT may be very sensitive for detection of early stage lung cancer, the costs of screening, the costs of the screening-incurred diagnostic evaluations, and the cost-effectiveness of such a screening program remain a serious concern.

The purpose of this article is to analyze the potential cost-effectiveness of lung cancer screening with CT from the perspective of national payer groups. Because it is a preliminary analysis, it requires several very strong assumptions, none of which has yet been proven. First and foremost, a cost-effectiveness analysis must assume that effectiveness exists. Therefore, we assume, for the purposes of this article, that screening for lung cancer with CT under a defined diagnostic protocol is effective; ie, there is a statistically significant lung cancer mortality rate reduction in a screened cohort compared to a control cohort. This has not yet been shown and would require a randomized controlled trial for definitive demonstration. Second, we assume that screening results in "down-staging" of detected cancers, ie, the stage distribution of cancers in the screened cohort is lower (earlier stage disease) than in the unscreened cohort. Third, we also assume that this down-staging results in a corresponding survival benefit as described in stage-specific survival rates. Finally, we must assume that the stage-specific survival rate benefits that occur from down-staging translate to mortality reduction benefits, ie, there are no effects of various biases (eg, lead time, length, overdiagnosis, stage migration, selection, historical control subjects).

Given these assumptions and the controversies surrounding the use of costly imaging technologies as screening tools, we impose fairly stringent conditions on our computations of the incremental cost/effectiveness ratio attributable to CT screening for lung cancer. As will be seen, we estimate for each component of this ratio a range of plausible values from one that is most favorable to the conclusion that CT screening is cost-effective to the one that is least favorable. We then sequentially always choose the least favorable figure, thus ensuring that the inferences we draw have met the most demanding standards. If this preliminary analysis of CT screening is positive, then we may proceed iteratively to test the assumptions of the analysis in a more rigorous controlled clinical trial.


   Materials and Methods
TOP
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
We estimated the ratios of expected costs to expected life-years for two hypothetical cohorts with lung cancer and monitored for a period of 15 years: (1) a screened cohort that receives five annual rounds of screening (one prevalence and four incidence rounds), and (2) an unscreened or usual care cohort. We emphasize that costs and life-years are expected values at diagnosis—forecasts of the cumulative health-care costs and annual survival probabilities of the average individual in each cohort over the follow-up period. We presuppose that these forecasts of cost and life expectancies will differ across the two cohorts, and that the difference between them, the incremental cost/life-year ratio, gauges the economic efficiency of the screening effort. Clearly, one way in which the health-care costs of the two cohorts differ is the cost of screening itself, which is zero for the unscreened cohort. More important is the likelihood that the costs of initial treatment and all continuing care, including more intensive care received in the last year of life, will differ between the cohorts. These differences will arise not only as a function of the stage at which the cancer is diagnosed, but also as a function of variations in the costs of different therapies. Of course, the factors that lead to cost variations also lead to differences in treatment outcomes, which we assume are fully reflected in survival differences. Thus, localized disease that can be treated effectively by surgery alone leads to better survival chances than late-stage disease. All other things equal, young and female patients have better prospects than older male patients. In principle, all such variations must be taken into account in estimating the incremental cost/life-year ratio attributable to screening.

In order to simplify the exposition, we summarize briefly how we computed the main elements of the incremental cost/life-year ratio in the following two subsections. As Siegel et al21 recommend, we set out a more detailed account of our methodology in a technical report that is available to interested readers upon request to the authors.

Screening and Screening-Incurred Costs
This cost component is estimated by considering a range of costs, utilization rates, and cancer yields for a defined screening and diagnostic protocol. The defined protocol for screening the hypothetical cohort is based on the protocol published by ELCAP.19 All subjects in the cohort received prevalence screening. All abnormal (positive result) screens generated incurred sequential diagnostic CT evaluations. Those subjects whose findings progressed during diagnostic CT evaluations underwent CT-guided percutaneous needle biopsy for diagnosis. Those subjects with cancer (either true-positive findings or interval cancers) did not receive subsequent screening, but rather received treatment and follow-up surveillance. Those subjects with negative screens or false-positive screen results who returned after diagnostic evaluation continued to receive subsequent rounds of screening.

As will be seen momentarily, we deal with the characteristics of the target population for screening in our estimation of treatment costs and life expectancy. The costs of screening are assumed to be invariant to the characteristics of the screened population. The number of subjects in the hypothetical cohort is also irrelevant, since the range of cost, utilization, and yield variables computes the expectations or conditional probabilities for the average person in the cohort. Total cost/effects for a cohort of given size are just scalars of these mean values.

Table 1 identifies the sources of data used for valuation of screening and diagnostic CT procedures. These values are computed using the reimbursement methodology of the federal Medicare program.


View this table:
[in this window]
[in a new window]

  		Table 1.. Costs of Screening and Diagnostic CT Procedures

 
Table 2 lists the "lowest" and "highest" values for the range of costs, utilization, and cancer yield rates considered in this analysis, as well as a "worst-case" scenario, which is comprised of highest costs and lowest yields for cancer detection. Cohort screening costs are computed for each set of variables. The "highest-cost, lowest-yield" screening cost is the figure then used for subsequent cost-effectiveness analysis. In order for us to be able to add this figure to the treatment cost estimates, we calculate the discounted expected value of screening costs by the means described below.


View this table:
[in this window]
[in a new window]

  		Table 2.. Parameters of Costs, Utilization, and Yields Used To Compute the Screening Component of Costs

 
The cost value used for screening CT represents the current Medicare global reimbursement for CT of the thorax without contrast material. These values are obtained by multiplying Medicare relative value units (RVUs) for a transitional facility, the relevant global reimbursement unit for professional/technical work, practice, and malpractice expenses, to the corresponding resource-based relative value scale conversion factor of $36.6137 for the year 2000. Since we adopt a national payer perspective for the analysis, we do not adjust for any geographic variations in the reimbursement computation.23 We assume that the unit cost of screening CT would be no higher than current global Medicare reimbursement, and likely would be much lower under high-volume, mass-screening conditions. The lowest cost of diagnostic CT represents the current global Medicare reimbursement for CT of the thorax with contrast material.23 The highest cost of diagnostic CT represents the current global Medicare reimbursement for CT of the thorax without and with contrast material.23 The lowest unit cost of CT-guided biopsy of the lung represents the current Medicare global reimbursement plus professional surgical fees for this entire procedure.23 The highest cost is 10 times the lowest, to account for potential additional costs of biopsy complications, such as hospital monitoring, nursing coverage, treatment of pneumothorax, and thoracotomies performed for benign findings.

In the ELCAP study, the rate of abnormal screening CT that required further diagnostic evaluation in the prevalence round of screening was 23%.19 Other ongoing screening trials, including one at our own institution, observe that the results of up to 50% of prevalence-screening CT examinations may be abnormal. Therefore, the range of abnormal prevalence-screening rates in this analysis is 25 to 50%; in the subsequent incidence-screening rounds, we have assumed up to 15% abnormal screening rates.

Screening periodicity is annual. For each abnormal screening CT finding, it is assumed that three additional diagnostic CT examinations will be incurred over 2 years to prove benignity of false-positive screen results (eg, at 3-, 6-, and 18-month intervals, in addition to the 12 month and 24 month screens), as outlined in the ELCAP protocol.19

The highest value for cancer prevalence (2.7%) is based on the ELCAP data.19 The lowest cancer incidence value is based on the earlier lung cancer screening trials9 10 11 12 13 14 15 16 17 18 with chest radiography and sputum cytology. Biopsy of benign and malignant findings is a significant cost determinant. The lowest value for biopsy utilization ratios are based on the ELCAP data: the total number of biopsy cases divided by the number of cancer cases, where the highest value assumes an equal number of benign and malignant biopsies.

As noted earlier, we always chose cost or utilization values that are likely to be as high or higher than those encountered in a true screening cohort. Our model assumes that all screening subjects receive all services based on the defined diagnostic protocol. The model then assumes the maximum costs in the screening cohort for the screening protocol under the assumptions given. The unscreened cohort would, in reality, also accrue the diagnostic costs of symptom evaluation. However, for the purposes of this analysis, the diagnostic costs of the unscreened cohort are assumed to be encompassed by the estimated costs of initial treatment. Since the same initial treatment costs estimates had to be used for the screened population, there is an unavoidable upward bias in the overall costs of the screened population. This bias, however, favors the "no-screening" option, and thus serves to make the cost-effectiveness of screening less, not more, likely.

Treatment Costs and Life-Years of Screened and Unscreened Cohorts
Table 3 summarizes the main assumptions used in estimating the treatment costs and life-years of both cohorts as well as the estimated values themselves, again divided between most favorable and least favorable to the hypothesis that screening is cost-effective. The salient aspects of these computations may be described briefly as follows:


View this table:
[in this window]
[in a new window]

  		Table 3.. Treatment Cost and Life-Year Computations

 
First, we model expected costs across the hypothetical 15-year follow-up by stage at diagnosis and the corresponding types of initial therapy for any person with a diagnosis of lung cancer. We piece together national estimates of the life-time treatment costs of lung cancer based on linked Medicare/Surveillance, Epidemiology, and End Results (program of the National Cancer Institute) records,24 and detailed information from the administrative/claims database and Cancer Registry of our center for this purpose. We begin by choosing a base cost figure and then key variations from that base attributable to differences in stage at diagnosis and initial treatment, as well as to the amount of time elapsed since diagnosis and the corresponding type of treatment at different time points thereafter—what some call phase of treatment. We use estimates of the cumulative cost of the initial 6-month of lung cancer treatment as this base amount. We then assign cost distributions across stage/treatment and phase as proportions of the base. Table 3 shows that we estimate the costs of treatment in the initial phase to be between $35,000 and $44,000, the smaller figure approximating the inflation-adjusted amount of initial treatment of a lung cancer case in the year 2000.

Our cost distributions are based in part on a fourfold classification of treatment phase: (1) the initial 6-month period after diagnosis, (2) the 12-month period immediately following the initial phase, (3) annual continuing care for all those surviving phase 2 up to the terminal period, and (4) terminal care provided during the final 18 months of life. These phase definitions roughly parallel those used to construct Medicare payment histories of lung cancer patients from diagnosis to death.24 We model the costs of each of these four phases as proportions of the estimated costs of the initial phase. Every cohort member incurs the costs of the initial period, whereas all other phase-specific costs are expected values obtained by multiplying the estimated, inflation-adjusted costs in year t by the probability that the cohort member is alive in year t.

We then divide each of the four phases by stage at diagnosis and initial course of therapy. Six categories are used for this purpose: local stage divided between surgery only and all other, more costly therapy combinations; regional stage divided between any single therapy and more costly multiple treatment modalities; and distal stage divided between any single and any multiple modality therapy. The estimated costs of phases 2 to 4 cross-classified by stage and the character of the first course of therapy are again taken in proportion to the base cost amount.

These proportional values are treated as constants in the analysis. However, we allow for variations in the percentages of the cohort with diagnoses at localized, regional and distal stages, as well as for the proportions of patients at each stage who receive less or more extensive treatment. These therapy subgroupings are based mostly on cost/utilization data from our center, eg, initial therapy for approximately 70% of our patients with localized disease is surgery alone; the remaining 30% receive various combination therapies such as surgery and radiation. As Table 3 shows, we assign these weights to the more favorable column, and we assume a random draw, ie, an even 50:50 split, between surgery alone and all combination therapies to the less favorable column, since in this case proportionately fewer patients would have the lowest cost treatment option. Similar logic explains the entries for the other two stage groupings.

Expected life-years are computed as the sum of conditional survival probabilities by single years of age for men and women in three age groupings over the synthetic follow-up period of 15 years. We use relative survival rates (RSRs) for cancer patients classified by stage at diagnosis, age group, and gender,25 and abridged life table values for the general population in the corresponding age group and gender categories26 to estimate these survival probabilities. The three age groupings (45 to 54 years, 55 to 64 years, and 65 to 74 years) were selected in part because published data on relative survival by stage used these specific brackets, and in part because persons of these ages would doubtless be the prime target population for lung cancer screening. We first conditioned annual probabilities of survival in the general population of American men and women on having survived to the base age of each age group (ie, 45 years in the group aged 45 to 54 years, 55 years in the group aged 55 to 64 years, etc.), thereby creating six series of conditional life table values for hypothetical follow-up periods of 15 years for each age/gender group. We then scaled the value of each year in these series by RSRs computed for each of the age/sex groups and further classified by stage at diagnosis and years since diagnosis to estimate the annual survival probabilities of lung cancer patients.26 We call attention to the entries in Table 3 that indicate the same RSRs and annual survival probabilities are used to estimate both less favorable and more favorable cost and life-year estimates. Survival outcomes are assumed to be uniquely associated with stage at diagnosis, age group, and gender, and these relationships remain constant over the follow-up period. What differs is the composition of the cohort in respect to these groupings, especially the fraction with diagnoses at earlier stages by virtue of CT screening.

Disaggregated estimates of expected costs and life-years are thus prepared for 18 subgroups delineated in respect to three diagnostic stages, three age categories, and gender. The computations for each subgroup, in effect, begin by estimating the annual survival probabilities for each year of the 15-year follow-up period. The cumulative sum of these probabilities represents the expected life-years of each subgroup, conditional on having survived to the respective base year of the subgroup. Estimated treatment costs for each phase and corresponding year in the follow-up are then scaled by the annual survival/death probability for that year. These expected costs are then adjusted for rising medical care price levels, and the opportunity costs of time, ie, the discount rate. Note that both the less and more favorable computations in Table 3 use an annual inflation rate of 5% and a discount rate of 7.5%. We describe later how the sensitivity of the results to these values was tested. The cumulative sum of these discounted expected costs yields the present value of the cost stream at the point of lung cancer diagnosis. It must also be noted that that we do not discount the life-year streams, ie, we treat all lives saved as being equal in value, irrespective of the point in the follow-up period at which the saving takes place. We call attention again to the technical report, available from the authors on request, which sets out in detail the data and formulas used to compute expected cost and life-year streams.

Finally, estimates for the average member of a given cohort are prepared by weighting each of the 18 stage/age/gender cumulative life-year and present value cost figures by the proportional representation of each subgroup in the overall cohort. Stage-related weights are always made explicitly, because it is primarily through changes in these proportions that the effects of screening are realized. In Table 3 , we use as the most favorable estimate the assumption that 20% of all lung cancers are diagnosed at the local stage and only 40% are diagnosed at the distal stage. The least favorable estimates assume more realistically that only 1 in 10 cancers is diagnosed at the local stage and half are distal. Age and gender proportions, however, can take many different values, depending on whether specific targeting occurs and/or whether chance elements play out so that, say, more men or more young persons, etc., receive a diagnosis of lung cancer. All other things being equal, a random draw of age and gender subgroups will result in lower costs and greater life expectancy than the age-specific and gender-specific proportions of lung cancer cases observed at the present time. Thus, we generate more favorable and less favorable estimates of costs and life-years by assuming respectively random draw (equal age group and gender weights) and observed differentials (current age group and gender ratios) in weighting subgroups comprising a given cohort. The sum of these annual probabilities then yields estimates of the life expectancies (remaining life-years) of individuals in each of the 18 subgroupings. Expected life-years of a target population are then estimated by weighting each of the 18 series by its proportion in that population.


   Results
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
Table 4 presents the main results of our analysis. Under "selected estimates," we present the plausible range of expected costs and life-year estimates based on the least favorable assumption set out in Tables 2 , 3 . Under these conditions, the average member of the no-screening or usual-care cohort would face the prospect of incurring expected costs on the order of $66,300 and have an average life-expectancy of just > 1 year. A screening effort that yields 70% of all detected cancers at the local stage (condition 1C) changes these prospects in two ways: first, the average cohort member now faces expected cumulative costs of only $65,280 and a life-expectancy of slightly > 39 months, results entirely of the down-staging over the cohort attributable to screening. Second, the expected cost of screening for the average member of the screening cohort must now be added to the treatment costs ($65,280 + $76,400 = $141,685). The incremental ratio of (extra) costs and life-years attributable to screening under these conditions is thus approximately $33,500. As conditions 1A and 1B suggest, the incremental cost per life-year ratio is quite sensitive to the fraction of the screened cohort that is detected with early stage disease. When this proportion is at 50%, the incremental ratio is $48,400, but when it is 30% it rises to > $90,000. We note here that these calculations assume that the entire cohort is screened. Clearly, if < 100% of the cohort is screened, the incremental cost/life-year ratio will increase somewhat.


View this table:
[in this window]
[in a new window]

  		Table 4.. Expected Costs and Life-Years of Lung Cancer Under Alternative Conditions and Screening

 
The "selected sensitivity tests" shown in Table 4 provide two points of comparison, one showing the sensitivity of the results to the set of most favorable assumptions, and the other to a specific assumption embedded in all of the other computations (a discount rate of 10% instead of 7.5%). In each case, we use the 50% local disease target as the comparator. Neither of these sensitivity tests, like a number of others we carried out but do not report here, yields unexpected results: more favorable assumptions about the age and sex characteristics of the cohort and the likelihood of more vs less extensive therapy reduces the incremental ratio by approximately $14,000. Raising the discount rate by 2.5 percentage points (which with the inflation rate of 5% is equivalent of increasing the net discount rate from approximately 2.4 to 4.8%) lowers the incremental ratio by approximately $3,000. Given that we have built these results mainly on the least favorable assumptions, the plausible range of the actual mean of the screened cohort are doubtless lower, perhaps to a substantial degree.


   Discussion
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
How do our cost-effectiveness figures stack up against others in the literature? In Table 5 , we set out some specific cost per life-year estimates for lung cancer screening as well as for other related cancer and noncancer screening and prevention strategies. For purposes of comparability, each of these estimates has been adjusted for the rate of general medical inflation up to the year 2000. The estimates for lung cancer are the only ones we found in the recent literature, and each is acknowledged by their authors to be relatively crude. The first is based on a mass screening effort in Japan in the early 1990s using annual chest radiographs.27 The comparatively high cost per life-year saved is due, in part, to the wide net cast by the effort at mass screening, and in part because the authors adjust for radiation-induced secondary malignancies arising from the screening effort itself. The second figure is in the nature of a "back-of-the envelope" calculation of using chest radiography to screen American smokers and former smokers for lung cancer.28 The author’s estimate is significantly lower than ours for several reasons. For one thing, he assumes implicitly that cancer survivors go back on the standard life table after their tumor is detected and treated, which substantially overestimates the life-year yield of the screening effort. For another, he does not enumerate the costs of false-positive results, which as we have seen substantially increases the cost of detecting a cancer.


View this table:
[in this window]
[in a new window]

  		Table 5.. Selected Cost Per Life-Year Comparisons

 
In Table 5 , we also show other cost per life-year estimates assembled recently in a wide-ranging compendium of cost-effectiveness results.29 These ratios were not selected in an entirely random fashion, but rather to provide focused comparisons with interventions/target populations that many agree are cost-effective, and have cost per life-year ratios in the same general range of our estimates. Thus, few would disagree in principle that preventive programs designed to get people to stop smoking are cost-effective. A recent study of nicotine gum as an aid in such efforts estimated an (inflation-adjusted) cost per life-year of approximately $13,000; hypertension screening at age 60 and screening donors for HIV have approximately the same cost/life-year ratios. Although we must be cautious in making these comparisons because of differences in methodology, the cost-effectiveness of many health-related interventions are frequently judged by their concordance with other such ratios. On these grounds, our estimates appear to be sufficiently reasonable and economically feasible to argue that cost factors should not be used to rule out more definitive controlled trials of the use of CT scans to screen for lung cancer.

Nonetheless, there are several limitations in interpreting our study. First, we have assumed that screening for lung cancer with CT is effective; ie, there is a statistically significant lung cancer mortality rate reduction in a screened cohort compared to a control cohort. This has not yet been proven and could require a randomized controlled trial for definitive demonstration. Second, we also assume that the down-staging that occurs with screening-detected cancers results in a corresponding stage-specific survival rate benefit that translates to mortality reduction benefits. This means we have not attempted to factor the effects of biases (eg, lead time, length, overdiagnosis, stage migration, selection, historical control subjects). Third, we have not tried to estimate the rate or costs of interval cancers. Instead, we have presented our results as a function of the proportion of cancers in the screened cohort that are detected at localized stage, assuming that any interval cancers would more likely reflect the stage distribution of cancers in the unscreened cohort. Fourth, and perhaps related to the previous item, we have not estimated the incremental costs nor effectiveness of other related screening tests. For example, molecular sputum markers, such as the up-regulated gene expression of heterogeneous nuclear ribonucleoprotein A2/B1, have been shown to be sensitive for preclinical detection of lung cancer.30 Sputum molecular markers detect central airway squamous cell neoplasms efficiently, and may be complementary to CT screening, which detects peripheral adenocarcinomas more easily. Addition of such additional screening tests could reduce the interval cancer rate of CT alone and improve the overall sensitivity of screening. Fifth, we have not included in this study the specialized costs of quality assurance or regulatory oversight in a screening program. We have observed such additional costs in breast cancer screening with mammography through the Mammography Quality Standards Act of 1992.31 The US General Accounting Office has estimated that such costs amount to an additional 3.4 to 4.2% per mammogram in the United States.32

The use of our synthetic cohort model is a potential limitation, as is the range of values and assumptions we have considered. The fact that we adopted a national payer perspective means that some social cost aspects such as patient opportunity costs have been omitted from the analysis. Moreover, we gauged only life-year gains, not quality-adjusted life-year gains. When data availability permits, adding these dimensions to the incremental cost-effective ratio may well change the results. However, a preliminary analysis such as this one may inform the current policy debate and help focus the research agenda in respect to lung cancer screening. When current screening trials accumulate and publish more data, and as new, possibly randomized trials, occur, the preliminary analysis can be refined and extended with better data.

However, we have not attempted to present an "optimized" screening program, ie, an ideal target population with high cancer prevalence and incidence, an overly sensitive or inexpensive screening test, minimal diagnostic evaluations, or limited surveillance. We have not attempted to examine cost-effectiveness of screening in specific groups with different risks, eg, current smokers, former smokers, or those with industrial exposure, etc. In each instance, we have tried to choose the most conservative option of screening or treatment costs, utilization rates, cancer yield, or outcomes. Stated differently, we have tried to construct a worst-case economic scenario for the hypothesis that lung cancer screening with CT is cost-effective.

Based on our analysis, and within the very conservative range of variables and assumptions that we have stated, if screening for lung cancer is effective, it is likely to be cost-effective if the screening process can detect > 50% of cancers at localized stage.


   Footnotes
 
Abbreviations: ELCAP = Early Lung Cancer Action Project; RSR = relative survival rate; RVU = relative value unit

Supported in part by an Advanced Cancer Detection Center grant, Department of Defense, Army DMD 17-98-8659.

Received for publication February 20, 2001. Accepted for publication December 12, 2001.


   References
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Jemal, A, Thomas, A, Murray, T, et al (2002) Cancer statistics 2002. CA Cancer J Clin 52,23-47[Abstract/Free Full Text]
  2. Fry, WA, Phillips, JL, Menck, HR (1999) Ten-year survey of lung cancer treatment and survival in hospitals in the United States: a National Cancer Database report. Cancer 86,1867-1876[CrossRef][ISI][Medline]
  3. Wagner, H, Ruckdeschel, JC (1996) Lung cancer. Reintgen, DS Clark, RA eds. Cancer screening ,118-149 Mosby St. Louis, MO.
  4. Rosell, R, Gomez-Codina, J, Camps, C, et al (1994) A randomized trial comparing preoperative chemotherapy plus surgery with surgery alone in patients with non-small cell lung cancer. N Engl J Med 330,153-158[Abstract/Free Full Text]
  5. Mountain, CF (1997) Revisions in the international system for staging lung cancer. Chest 111,1710-1717[Abstract/Free Full Text]
  6. Inoue, K, Sato, M, Fujimura, S, et al (1998) Prognostic assessment of 1310 patients with non-small cell lung cancer who underwent complete resection from 1980 to 1993; Thorac Cardiovasc Surg 116,407-411
  7. Smith, RA, Mettlin, CJ, Davis, KJ, et al (2000) American Cancer Society guidelines for the early detection of cancer. CA Cancer J Clin 50,34-49[Abstract]
  8. Kurie, JM, Spitz, MR, Hong, WK (1995) Lung cancer chemoprevention: targeting former rather than current smokers. Cancer Prev Int 2,55-64
  9. Sanderson, DR (1986) Lung cancer screening: the Mayo study. Chest 89(suppl),324S[Medline]
  10. Tockman, M (1986) Survival and mortality from lung cancer in a screened population. Chest 89(suppl),324S
  11. Martini, N (1986) Results of the Memorial Sloan-Kettering study in screening for early lung cancer. Chest 89(suppl),325S[Free Full Text]
  12. Kubik, A, Parkin, DM, Khlat, M, et al (1990) Lack of benefit from semi-annual screening for cancer of the lung: follow-up report of a randomized controlled trial on a population of high-risk males in Czechoslovakia. Int J Cancer 45,26-33[ISI][Medline]
  13. Fontana, RS, Sanderson, DR, Taylor, WF, et al (1984) Early lung cancer detection: results of the initial (prevalence) radiologic and cytologic screening in the Mayo Clinic study. Am Rev Respir Dis 130,561-565[ISI][Medline]
  14. Sanderson, DR (1989) Lung cancer: the American view. Eur Respir J 2,1002-1007[Abstract]
  15. Fontana, RS, Sanderson, Dr, Woolner, LB, et al (1991) Screening for lung cancer: a critique of the Mayo Lung Project. Cancer 67,1155-1164[CrossRef][ISI][Medline]
  16. Miller, A (1986) Lung cancer screening: summary. Chest 89(suppl),325S
  17. Eddy, D (1989) Screening for lung cancer. Ann Intern Med 111,232-237
  18. Stitik, FP, Tockman, MS (1978) Radiographic screening in the early detection of lung cancer. Radiol Clin North Am 16,347-366[ISI][Medline]
  19. Henschke, CI, McCauley, DI, Yankelevitz, DF, et al (1999) Early Lung Cancer Action Project: overall design and findings from baseline screening. Lancet 354,99-105[CrossRef][ISI][Medline]
  20. Kaneko, M, Eguchi, K, Ohmatsu, H, et al (1996) Peripheral lung cancer: screening and detection with low-dose spiral CT versus radiography. Radiology 201,789-802
  21. Siegel, JE, Weinstein, MC, Torrance, GW (1996) Reporting costeffectiveness studies and results. Gold, MR Siegel, JE Russell, LBet al eds. Cost effectiveness in health and medicine ,276-303 Oxford University Press New York, NY.
  22. Current procedural terminology. 4th ed. 1999 American Medical Association Chicago, IL.
  23. 64 Federal Register 211 (1999). Health Care Financing Agency, US Dept Health and Human Services
  24. Riley, GF, Potosky, AL, Lubitz, JD, et al (1995) Medicare payments from diagnosis to death for elderly cancer patients by stage of diagnosis. Med Care 33,828-841[CrossRef][ISI][Medline]
  25. Wingo, PA, Gloeckler, Ries LA, Parker, SL, et al (1998) Long-term cancer patient survival in the United States. Cancer Epidemiol Biomarkers Prev 7,271-282[Abstract]
  26. Anderson RN. United States Life Tables, 1997. National Vital Statistics Reports, 1999; 47(28)
  27. Baba, Y, Takahashi, M, Tominguchi, S, et al (1998) Cost-effectiveness decision analysis of mass screening for lung cancer. Acad Radiol 5(suppl2),S344-S346
  28. Strauss, GM (1999) Screening for lung cancer: an evidence-based synthesis. Surg Oncol Clin N Amer 8,747-774
  29. Tengs, TO, Adams, ME, Pliskin, JS, et al (1995) Five-hundred life-saving interventions and their cost-effectiveness. Risk Anal 15,369-390[CrossRef][ISI][Medline]
  30. Tockman, MS, Mulshine, JL, Piantadosi, S, et al (1997) Prospective detection of preclinical lung cancer: results from two studies of heterogeneous nuclear ribonucleoprotein A2/B1 overexpression. Clin Cancer Res 3,2237-2246[Abstract/Free Full Text]
  31. 62 Federal Register 208 (1997). Food and Drug Administration, US Dept. Health and Human Services
  32. Cost-benefit analysis: memorandum B278642GAO, GAO/OGC-98-14. 1997 United States General Accounting Office Washington, DC.



This article has been cited by other articles:


Home page
 NEJMHome page
The International Early Lung Cancer Action Program
Survival of Patients with Stage I Lung Cancer Detected on CT Screening
N. Engl. J. Med., October 26, 2006; 355(17): 1763 - 1771.
[Abstract] [Full Text] [PDF]

Home page
 Ann OncolHome page
P. Das, A. K. Ng, C. C. Earle, P. M. Mauch, and K. M. Kuntz
Computed tomography screening for lung cancer in Hodgkin's lymphoma survivors: decision analysis and cost-effectiveness analysis
Ann. Onc., May 1, 2006; 17(5): 785 - 793.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Epidemiol. Biomarkers Prev.Home page
J. E. Lafata, J. Simpkins, L. Lamerato, L. Poisson, G. Divine, and C. C. Johnson
The Economic Impact of False-Positive Cancer Screens
Cancer Epidemiol. Biomarkers Prev., December 1, 2004; 13(12): 2126 - 2132.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Respir. Crit. Care Med.Home page
A. McWilliams, J. Mayo, S. MacDonald, J. C. leRiche, B. Palcic, E. Szabo, and S. Lam
Lung Cancer Screening: A Different Paradigm
Am. J. Respir. Crit. Care Med., November 15, 2003; 168(10): 1167 - 1173.
[Abstract] [Full Text] [PDF]

Home page
 JAMAHome page
T. N. Chirikos, T. Hazelton, M. Tockman, and R. Clark
Cost-effectiveness of Screening for Lung Cancer
JAMA, May 14, 2003; 289(18): 2358 - 2358.
[Full Text] [PDF]

Home page
 JAMAHome page
P. J. Mahadevia and N. R. Powe
Cost-effectiveness of Screening for Lung Cancer--Reply
JAMA, May 14, 2003; 289(18): 2358 - 2359.
[Full Text] [PDF]

Home page
 ChestHome page
J. Blum, N. Rinne, M. K. Gould, T. N. Chirikos, T. Hazelton, M. Tockman, and R. Clark
Lung Cancer Screening Debate
Chest, February 1, 2003; 123(2): 653 - 655.
[Full Text] [PDF]

Home page
 Am. J. Respir. Crit. Care Med.Home page
S. G. Spiro and J. C. Porter
Lung Cancer--Where Are We Today?: Current Advances in Staging and Nonsurgical Treatment
Am. J. Respir. Crit. Care Med., November 1, 2002; 166(9): 1166 - 1196.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (24)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chirikos, T. N.
Right arrow Articles by Clark, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chirikos, T. N.
Right arrow Articles by Clark, R.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS