SEARCH

SEARCH BY CITATION

Keywords:

  • high risk;
  • low-dose computed tomography;
  • noncalcified nodules;
  • overdiagnosis bias;
  • screening;
  • smokers

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

While low-dose CT scans have been shown to detect greater numbers of early lung cancers than conventional CXR, the first randomized trial of CT versus CXR screening in more than 50 000 subjects has shown a 20% reduction in mortality with CT. There are several other randomized trials in progress. CT scanning may be a useful technique for identifying lung cancer at an earlier stage and may reduce mortality. However, before it can be used on a wider scale, issues such as overdiagnosis bias, cost-effectiveness, false positive findings of multiple noncalcified nodules and the willingness of the relevant population to accept CT scanning need to be evaluated. There is still very little information on the cost per life-year saved as a result of CT scanning, as the data to date is very imprecise. There is no evidence that screening programs influence smoking rates despite the inclusion of cessation programs in many trials. Furthermore, if CT screening is adopted, much work is needed to persuade individuals at high risk, mostly current or former heavy smokers with some airflow obstruction, to participate in a screening program.


INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

Lung cancer remains the most challenging of all the malignant diseases. It causes the greatest mortality among men and women dying of any cancer across the world. Lung cancer rates in never-smokers have risen so that it is now included in the top 10 causes of death from cancer in the Western world; yet smoking, the primary cause of lung cancer in 80% of sufferers, is being countered with smoking cessation programs that have had variable success across the globe. Rates of smoking, especially among young women, continue to increase.

Studies evaluating screening for lung cancer were completed 30 years ago. These studies were based on annual CXR screens, with or without additional sputum cytological examinations. There were five studies in all, three in the USA, one in London and one in Czechoslovakia.1–5 These studies all showed that a greater number of cancers were detected in the screened group and more of these patients underwent surgical resections, but there was no long-term benefit in terms of overall mortality. Thus, a promising idea was not pursued further.

Since then, efforts have concentrated on treatment, with the advent of safer thoracic surgery, improved staging, and better and more focused radiotherapy, and there has been a great effort to improve palliation in patients with advanced disease. Nevertheless, progress has been very slow. There has been virtually no improvement in the outlook for patients with small-cell lung cancer during the last 20 years. Response rates to chemotherapy in patients with advanced non-small-cell lung cancer (NSCLC) have reached a pretty dismal plateau, and the focus is now on targeted therapies aimed at identifying and treating patients with particular mutations and genetic depletions. While it is vital to maintain and, if possible, increase the momentum for treating the tens of thousands of patients who present with lung cancer, for most of those with advanced, incurable disease, the trend is swinging back to investigation of the ‘front end ’ of the disease.

Lung cancer is, in many ways, a ‘silent killer’, which in more than 70% of new cases, presents too late for curative treatment to be attempted. The lung contains no pain fibres, and masses may grow to a considerable size before symptoms become apparent. There are no respiratory symptoms or signs that are specific to lung cancer; suggestive symptoms such as cough, wheeze, chest pain and even haemoptysis each have much commoner causes. The average family practitioner in the UK will see between one and two new cases of lung cancer in a year among hundreds of patients with coughs and respiratory infections, making the early identification of a new lung cancer almost impossible. Paradoxically, and not surprisingly, those patients whose tumours are identified by chance tend to have the best prognosis after treatment.

LOW-DOSE CT SCREENING FOR LUNG CANCER

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

The greater sensitivity of CT scanning has rekindled the concept of early detection that was proposed some 10 years ago and has lead to several hypothesis-generating studies across the world. Most of these studies confirmed that a higher percentage of lung cancers can be identified using low-dose spiral CT, as compared with CXR; more cancers were detected in early stages of disease (IA and IB), and in the main, these were resectable. These observations have resulted in the establishment of randomized controlled trials (RCT) of CT screening versus either standard CXR or usual care (Table 1) to determine whether screening by low-dose spiral CT, in particular, can reduce mortality due to lung cancer. Preliminary findings from the first of these studies, the American National Lung Cancer Screening Trial (NLST), were reported in June 2011.6 This trial showed that screening for lung cancer using low-dose CT reduced mortality due to lung cancer by 20% and all cause mortality by 6.7%, as compared with CXR screening. The trial included 53 454 persons who were enrolled between August 2002 and April 2004; all were high-risk, 30-pack-year smokers who were recruited at 33 US medical centres. Of this total, 26 722 participants were randomized to three annual screens by low-dose CT, whereas the 26 732 participants in the control arm underwent three annual screens by single-view CXR. The percentage of individuals with positive screens was high: 24.2% in the CT arm and 6.9% in the CXR arm. Of these, 96.4% of the positive screens in the CT arm and 94.5% of the positive screens in the CXR arm were false positive results due mainly to the detection of benign noncalcified nodules (NCN). The incidence of lung cancer was 645 cases per 100 000 person years (a total of 1060 cancers) in the low-dose CT group compared with 572 cases per 100 000 person years (941 cancers) in the CXR group. There were 247 deaths due to lung cancer per 100 000 person years in the CT group, as compared with 309 in the CXR group representing a relative reduction in deaths due to lung cancer of 20%. This result, although very important and likely to receive considerable attention from both the medical profession and general public, was derived from a single, albeit very large, trial. Other trials of similar but not identical design are in progress and are briefly summarized here.

Table 1.  Randomized controlled trials of CT for lung cancer screening
TrialRecruitment periodTarget population and smoking historySample sizeIntervention armControl armResults
  1. FOB, fluorescence fibre-optic bronchoscopy; LDCT, low-dose CT; LLP, Liverpool Lung Project; NLST, National Lung Cancer Screening Trial; UKLS, UK Lung Screen.

NLST2002–2004Age 55–74 years, with ≥30 pack years53 454LDCT at 1, 2 and 3 yearsCXR at 1, 2 and 3 yearsReduction in lung cancer mortality of 20%. Rate of death from any cause was reduced by 6.7%
NELSON2003–2006Age 50–75 years, with >15 pack years16 000LDCT at 1, 2 and 4 yearsUsual careExpected 2016
DLCST2004–2006Age 50–70 years, with >20 pack years4104LDCT at 1, 2, 3, 4 and 5 yearsUsual careExpected 2016
DANTE2001–2006Men only, age 60–74 years, with ≥20 pack years2472LDCT at 1, 2, 3, 4 and 5 yearsCXR at baseline, followed by usual careAfter median follow-up of 33 months, lung cancer mortality was 1.6% in LDCT group and 1.7% in control group
ITALUNG2004–2006Age 55–69 years, with ≥20 pack years3206LDCT at 1, 2, 3 and 4 yearsUsual careNot reported
MILD2005–ongoingAge 50–75 years, with ≥20 pack yearsUnknownLDCT (randomized to either annual or every 2 years)Usual careNot reported
LUSI2007–ongoingAge 50–69 years, with ‘heavy’ smoking history4000LDCT at 1, 2, 3, 4 and 5 yearsUsual careExpected 2015–2017
DEPISCAN2002–2004Age 50–75 years, with ≥15 pack years765LDCT at 1, 2 and 3 yearsCXRNot reported
Lung SEARCH2008–2011Age 50–75 years, with >20 pack years and mild/moderate COPD1546Annual sputum cytology/cytometry at years 1–5. FOB/CT if abnormalCXR at end of 5 yearsNot reported. Will be completed 2016
UKLS2011–ongoing5% risk of developing lung cancer in 5 years using the LLP risk prediction model4000 for pilot study, with intention to recruit 32 000 in totalSingle LDCTUsual careExpected 2013

In the Italian DANTE study,7 2472 men aged between 60 and 74 years were recruited. They had a 20-plus pack year smoking history, and if they had quit, it was within the last 10 years (in fact, the subjects had smoked for a mean of 47 pack years). All underwent CXR and sputum cytological assessment at baseline and were then randomized to an initial CT with four annual follow-ups, or annual clinical examinations for the control group. Participants were recruited through local general practitioners and by advertising. Among the CT group, 28 cancers were detected, of which 13 were visible on the prerandomization CXR. This gave a prevalence rate of 2.2%. Sixteen of these tumours were stage I, and 68% were resectable. Among subjects in the control arm, eight cancers were detected at baseline (a prevalence of 0.67%), of which four were stage I and six were resectable. Adenocarcinoma and bronchoalveolar cell carcinoma (BAC) accounted for 61% of the cancers in the CT arm and 50% of the cancers in the control arm. Although, as in most studies, the CT arm contained more patients with low-stage and resectable disease, the effects of the type of screening on mortality have not yet been reported.

For the French DEPISCAN study,8 1000 men and women attending general practices were enrolled. They were asymptomatic, aged between 50 and 75 years, and had smoked more than 15 cigarettes per day for 20 years or had quit within the last 15 years. They were randomized to two annual screens by low-dose CT or CXR. Complete data was available for only 621 subjects. Eight lung cancers were detected among subjects in the CT-screening arm, but five of these were at an advanced stage (IIIb or IV) and only one was at stage IA. This compared with just the one lung cancer detected among subjects in the CXR arm. Moreover, 45% of subjects in the CT arm had abnormal scans, as compared with 7.6% of subjects in the control group.

The NELSON trial9 is a Belgian-Dutch collaborative trial involving 15 428 subjects (as of October 2005) who were randomized to CT at baseline and at 1, 2 and 4 years, as compared with a control arm that received no screening. Potential participants were initially identified by a questionnaire sent to 335 441 individuals, mainly men aged 55–75 years who had smoked at least 15 cigarettes per day for 25 years or 10 cigarettes per day for 30 years, or had quit smoking within the last 10 years. The results from this trial are not yet available, but selection was based on degree of risk for lung cancer, and it was calculated that 28 000 participants would have been required to demonstrate a reduction in lung cancer mortality of 20%, whereas for the NLST, it had been predicted that 50 000 participants would be required to demonstrate a difference of this size. The NELSON trial was, however, designed to detect a difference in mortality of 25% after 10 years of follow-up. The authors also hope to combine this data with that from another Danish trial in which 4000 subjects have been enrolled. The NELSON study also has a control arm with no CXR screening, but subjects over 70 years of age were not included; the participants smoked more heavily, and there were more females (15% in the NELSON study and 45% in the Danish study). However, analysis of other prospective cohort studies showed that there were no differences in mortality due to lung cancer between the genders.10 The results of baseline screening for the NELSON trial are available, and 70 cancers were detected, with another 20 being detected to date; three were interval cancers, and 17 were identified at 1 year, as nodules detected during prevalence screening increased in size.

The ITALUNG cohort comprises 3206 subjects selected from Italian general practices.11 Participants were randomized to 4 years of annual CT screening or usual care. They were between 55 and 69 years of age, had a smoking history of at least a 20 pack years and if they had quit, had done so within the last 10 years. Among participants in the CT arm, 21 cancers were detected during prevalence screening (a rate of 1.5%), with 10 being at stage I and 16 patients undergoing pulmonary resections. In all, 30% of the CT scans were abnormal mainly due to the presence of NCN.

This data was similar to that from an initial RCT in the USA, which was performed as a feasibility study leading up to the major NLST. Gohagan et al.12 performed an RCT of low-dose CT versus CXR, with 1660 subjects in each arm. The yield of cancers at baseline was 1.6% for CT and 0.45% for CXR but decreased to 0.57% for CT at the first incidence screen at 1 year. Forty lung cancers were diagnosed by CT over the study period, and 20 were diagnosed by CXR, with 48% and 40%, respectively, being at stage I. There were 16 and 9 stages III and IV cancers in the two arms, respectively, hinting at a stage shift.

In another study, the UK Lung Screen (UKLS), that is about to begin in the UK,13 4000 volunteers will be enrolled for randomization to single CT screening versus usual care. This study will assess whether early diagnosis reduces mortality and whether the benefits outweigh potential harm in a cost-effective manner. If initial findings are positive, another 28 000 participants will be enrolled. The design of the trial is based on the selection of subjects who are at high risk for lung cancer (5% over the 5 years of observation) and on the use of a validated risk identification model, the Liverpool Lung Project risk model. Only subjects with clearly defined abnormalities on their CT scans will be asked to undergo further scans depending on the analysis of nodule volumes, based on the nodule analysis scheme being used in the NELSON trial,14 with additional detailed assessment of tiny nodules. This novel design is based on minimizing exposure to radiation during CT, is less dependent on compliance with annual scans and will deliver a result within 5 years, which is a shorter time frame compared with other RCT of lung cancer screening.

The NLST and other trials that are in progress raise specific and general issues that need to be addressed when evaluating the effects of a screening test. These include:

  • • 
    overdiagnosis bias in screening trials
  • • 
    whether the NLST reveals a stage shift in the presentation of the cancers detected by screening, an important issue in a positive screening study
  • • 
    the effects of selection of the at-risk populations for the study
  • • 
    any additional value of a smoking cessation program, if it is included in the study design
  • • 
    the acceptability of a screening study to the high-risk, heavy-smoking or ex-smoking population
  • • 
    whether the cancers detected at screening are ‘different’ to those that present in routine clinical practice
  • • 
    whether the high numbers of benign NCN so frequently detected in CT-based studies has a confounding effect
  • • 
    whether there is a healthy volunteer effect
  • • 
    whether these screening strategies are cost-effective in an elderly population with a high prevalence of comorbidities; and finally
  • • 
    whether there are any long-term effects of exposure to radiation

For a screening test to be acceptable, the target disease should represent a significant health problem; it should be detectable in those who are at risk before symptoms occur. Detection should result in improved outcomes, the test should be acceptable to the target population, and it should only minimally detect pseudodisease. Lung cancer would certainly fulfil most of these criteria; however, there are potential problems both with recruitment of the target population and the incidence of false positive findings (see later).

BIAS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

There are, in particular, two important biases related to screened populations: lead-time bias and overdiagnosis bias. Lead-time bias refers to the apparently longer survival of many patients with cancers discovered at screening; this is due to early detection when the cancer is still asymptomatic. Patients will therefore survive longer than if they had been diagnosed later in the natural history of the disease, when they would present with symptoms, as occurs in control groups. This bias does not affect mortality, which is the ultimate end-point of a screening trial.

Overdiagnosis bias is the term applied to cancers, which are often very slow-growing, that are detected at screening but would otherwise have remained occult and would either have regressed or not caused the death of the patient. In a randomized screening trial, the numbers of cancers detected in each arm should ultimately be equal. If not, it strongly suggests that overdiagnosis has occurred in one arm, usually the screening arm. It may take 10–15 years for the detection rates in the two arms to equalize. Some researchers believe that many of the lung cancers detected by screening represent the slowest growing tumours. Certainly, small-cell lung cancers, which are the fastest growing type, are rarely detected during prevalence and, in particular, incidence screening. When they are detected, the proportions are less, relative to the usual distribution of cell types. It has been suggested that lung cancers detected due to overdiagnosis bias will have a volume doubling time (VDT) of >400 days.15 In a review of 1520 high-risk subjects screened in the 5-year Mayo Clinic screening program, the VDT of tumours that were imaged more than once was calculated. Among the 1520 subjects screened, 61 lung cancers were detected in 59 individuals. VDT was calculated in 49 cases, and the mean value was 518 ± 1049 days. In 27 of these cases, the VDT was >400 days, and most of these were adenocarcinomas. The mean VDT was longer in women (688 days) than men (234 days), and this was consistent across all tumour types. The authors concluded that overdiagnosis bias may occur, especially in women.16

Current trials are investigating nodules detected during prevalence screening based on their growth or VDT, with 400 days being the critical cut-off value. Nodules with VDT of <400 days and that are at least partly solid will be referred for direct assessment.13,14 Early evidence from 53 participants in the Danish Lung Cancer Screening Trial suggests that the combination of VDT analysis with assessment by PET may further improve the sensitivity and specificity for detection of malignant nodules at screening.17

STAGE SHIFT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

It follows that if it is mainly slow-growing lung cancers that are discovered at screening, these are likely to be smaller tumours representing ‘early’ lung cancer, and that the surgical stage is likely to be correspondingly low, that is, IA or IB. If this is so, it will influence survival data due to the natural prolonging of survival and ideally, result in a reduction in mortality, that is, a genuine improvement in survival among this elite group taking into account the problem of overdiagnosis bias. By the same token, early discovery of lung cancers in the screened population, as compared with a control group, should result in the identification of fewer cases of advanced disease and therefore induce a stage shift in the distribution of cancers towards those with lower staging. This hypothesis has been tested in a number of nonrandomized screening trials in various populations with differing risks of lung cancer.18–25 These studies reported a wide range of percentages of patients with stage I lung cancer (27–100%), and the most frequent presentation was at this stage. These trials also showed that whereas CXR screening discovered lung cancers in 0.5% of subjects screened, the percentage was about four times greater with low-dose CT, a prevalence of 2.7% in the New York 1000 Early Lung Cancer Action Project (ELCAP) volunteer study being the highest to date.18

In the NLST RCT, there were about three times as many cancers detected by screening among the low-dose CT group (649), as compared with the CXR group (279), a proportion that was similar to those in the single-arm nonrandomized CT-screening studies described previously. In the low-dose CT arm, 44 cancers were also discovered after a negative screening result, and 367 were discovered among participants who either missed their screening test or were diagnosed during the post-screening follow-up period. In the CXR group, 137 additional cancers were detected after a negative screening result, and 525 were detected among those who either missed their screening test or were diagnosed during follow-up. These large numbers of additional cancers confirm the lower sensitivity of CXR.

The staging and histological distribution of lung cancers detected in the screened and nonscreened populations in the NLST are summarized in Table 2. In total, 520 (50%) patients diagnosed with lung cancer in the low-dose CT arm of the trial had stage IA or IB disease. This compared with 289 (31%) patients with lung cancer in the control arm who had stage IA or IB disease. These data would appear to suggest that there was a stage shift towards a lower stage at presentation, which may be the reason for the lower mortality among the low-dose CT group. There was a preponderance of bronchioloalveolar carcinomas in the screened population, and small preponderances of adenocarcinoma and squamous cell carcinoma but not large cell, unspecified NSCLC or small-cell lung cancer. The stage shift observed in the NLST is seen to a lesser extent in preliminary reports on the other RCT that is still in progress; however, all these studies are considerably smaller than the NLST, and a meta-analysis approach will be required to confirm this effect.

Table 2.  Characteristics of lung cancers detected in the National Lung Screening Trial6
Stage and histological typeLDCT screening group (n = 26 722)CXR screening group (n = 26 732)
Lung cancers detected by LDCT screening (n = 649)Lung cancers detected after negative screening result (n = 411)Lung cancers detected by CXR screening (n = 279)Lung cancers detected after negative screening result (n = 662)
  • Adapted with permission from table 5 in Aberle et al.6

  •  

    Staging data was unavailable for 14 patients after a positive screening test and for six patients after a negative screening test in the LDCT group. In the CXR screening group, staging information was unavailable for four cancers after a positive CXR and for eight after a negative CXR.

  •  

    Includes bronchioloalveolar carcinoma.

  • LDCT, low-dose CT.

Stage, n (%)    
 I400 (63)120 (30)131 (48)158 (24)
 II46 (7)27 (6)25 (9)49 (7)
 III108 (17)113 (28)62 (23)169 (26)
 IV81 (13)145 (36)57 (21)278 (43)
Histology, n (%)    
 Adenocarcinoma353 (55)137 (34)125 (45)238 (36)
 Squamous cell carcinoma136 (21)107 (27)70 (25)136 (21)
 Large-cell carcinoma28 (4)13 (3)12 (4)31 (5)
 Non-small-cell carcinoma, not otherwise specified and other75 (11.6)56 (14)40 (15)118 (18)
 Small-cell carcinoma49 (7.6)88 (22)28 (10)131 (20)
 Carcinoid5 (0.8)1 (0)1 (0)1 (0)

CHOICE OF AT-RISK POPULATION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

The choice of study group will influence the cost-effectiveness of screening, as more cancers will be detected among higher risk groups. The variables that will increase the potential yield when screening for lung cancers are increased age, greater numbers of pack years smoked, a preponderance of male smokers—although this will change with more women now smoking in many countries—lower socioeconomic class and the presence of airways obstruction, with patients with severe COPD having the highest incidence and risk of lung cancer.26 Few studies have used FEV1 as a risk indicator, although the UKLS factors this into the Liverpool Lung Project risk model. There is a potential danger that with the publication of results from the NLST, individuals with a much lower risk, who are well but ‘worried’, will demand CT screening.

SMOKING CESSATION PROGRAMS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

Lung cancer screening trials offer an opportunity for providing advice on smoking cessation, and most trials include this. To date, there has only been one report from the NELSON trial, which indicated that the inclusion of such a program within the trial failed to show a beneficial effect.27 Two cohorts of male smokers, one comprising 550 subjects with negative screening results and the other comprising 440 subjects with indeterminate results, were sent questionnaires 2 years after entry into the trial. Although those with indeterminate results recorded a higher percentage of attempts to quit than the group with negative results, the difference was not statistically significant.

THE ACCEPTABILITY OF SCREENING

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

The population with lung cancer comprises mainly elderly individuals of lower socioeconomic status who often have significant comorbidities and is still mainly male. It is not obvious that this is a population that would be enthusiastic about screening. Many have taken a risk by smoking, are therefore not averse to taking risks and are not interested in their longer term health. Therefore, it is possible that a screening program that should one be set up would not attract adequate numbers of appropriate individuals.

Silvestri et al.28 performed telephone interviews with 2001 smokers, ex-smokers and non-smokers in the USA. Smokers were significantly more likely to be males, non-White, less educated to report a poor health status and history of cancer, and were significantly less likely to be regular users of health-care resources. In making the decision to undergo screening, non-smokers felt that the risk that the disease would be detected and the accuracy of the test were of greater importance than smokers did. Smokers were also less willing to pay for a screening test and were less willing to undergo treatment should lung cancer be detected. Smokers were also less willing to undergo any screening test, as compared with ex-smokers and never-smokers.

Among the CT-screening studies that provided information on how subjects were recruited, most indicated that only a minority of eligible subjects agreed to participate in the study.

The NLST recruited widely across the USA at 33 medical centres. Potential subjects were recruited through articles in the press, by local mailing, by advertising and through the Internet. Care was taken to recruit subjects from minority communities, but no information is available on the relative success of the campaign, that is, how many subjects did not wish to participate. The study population was, however, representative of the high-risk smoking population in the USA.29 In the NELSON study, a questionnaire was sent to 335 441 men aged 50–75 years, whose names were in a database of population registries.9 Of these, 106 931 replied, and subjects were selected on the basis of their smoking habits and risk factors, so as to minimize the number of participants needed. Of these, 11 103 (1.6% females) gave consent to participate in the study. This represented 3.3% of all the individuals initially approached. In a second round of recruitment in 2005, 250 000 questionnaires were sent out, and 44 509 individuals replied. Of these, 4535 have been randomized, representing 1.8% of the initial population that was approached. In the ITALUNG trial, a total of 3206 subjects were enrolled out of a total of 71 232 who were sent letters from 269 general practices.11 Again, the uptake was low, with 4.5% of all subjects approached being enrolled. In the DEPISCAN trial, 765 subjects were recruited from 205 general practices and by 25 occupational physicians, representing a median of six subjects from each active centre; however, only 41% of centres became active and were able to recruit subjects.8

In an ongoing UK study (Lung SEARCH), 1600 high-risk heavy smokers with mild/moderate COPD have been randomized to 5 years of annual sputum cytology/cytometry assessments with autofluorescence bronchoscopy and CT if there are abnormal findings versus 5 years of usual care with CXR when exiting the study. A qualitative study was performed to investigate the attitudes of subjects who (i) entered the study and had normal cytology results, (ii) those with abnormal cytology results who were undergoing further planned investigations, and (iii) a third group who refused to participate. The data shows that most subjects participated for altruistic reasons, whereas most of those who did not participate felt that they were too old, or were risk averse and not interested.

ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

Cancers detected by screening do differ with respect to the distribution of cell types identified at clinical presentation. More screening cancers are adenocarcinomas and, in particular, BAC. VDT was calculated for many of the stage I cancers detected by CXR in the American screening studies, and there was no evidence that these tumours had a significantly longer VDT than ‘normal’; indeed, only a handful of tumours had a VDT of >400 days, suggesting that there was no overdiagnosis bias in these studies,15 and, ultimately, that there were no differences in longer term mortality.

A recent retrospective Japanese study of never-smokers, whose cancers were detected by CT screening (n = 218) and CXR screening (n = 160), or who were symptomatic (n = 82), highlighted differences in cell type distribution.30 As might be expected, the type of cancer most frequently detected in the CT-screening group was BAC: 65% (142 cases) compared with 12% and 6% for CXR screening and symptomatic subjects, respectively. Other adenocarcinomas were detected in 32% of CT-screening patients, and in 72% and 74% of CXR screening and symptomatic subjects, respectively. There were very few cancers of other cell types detected in these populations, and the frequency of detection was least in the CT-screening group, as CT is much better for detecting peripheral BAC and adenocarcinomas rather than the more central, squamous and small-cell tumours. Five-year survival was much greater in the patients receiving CT screening and was worst in those presenting with symptoms. Although these results are encouraging, lead-time bias must be a significant confounding factor.

It could be argued that if early cancers are detected by screening, then the number of advanced tumours detected should decrease, that is, there would be stage shift. However the natural history of adenocarcinomas of the lung has never been investigated, and Bach31 has suggested that there are two types of lung cancers: early cancers detected by screening and advanced tumours that present late and cause death. In the earlier screening trials that used CXR and sputum cytology, twice the number of early cancers were detected in the screening arms, whereas equivalent numbers of advanced cancers were detected in the two study arms, leading Bach to suggest that these tumours showed a bimodal distribution pattern.31

The detection of even earlier cancers by CT than by CXR should be a good test of the bimodal model, that is, there should be a better ‘dose–response’ with more earlier cancers and fewer advanced cancers detected rather than more earlier tumours and the same percentages of patients with advanced disease. Bach et al.32 assessed the results from three nonrandomized CT-screening studies that included more than 3000 individuals. Of these, 144 were diagnosed with lung cancer, and 67% of these cancers were at an early stage. They found 10 times more early lung cancers than seen with CXR. Many more individuals went on to have a lung resection than predicted by validated modelling of expected numbers for a similar population. However, the number of advanced cancers detected was identical to that predicted for this population, and the number of deaths was also similar. Furthermore, in the 3–9 years of follow-up, there was no evidence of a reduction in the numbers of advanced cancers diagnosed. Another observation from this study was that early cancers were not as closely associated with the usual risk factors of heavy smoking and male gender. Henschke et al. performed a CT-based trial and showed that more early cancers occurred in women, and this was not closely related to smoking history.33 This finding is also consistent with autopsy data, which showed that more early, indolent cancers were detected in women, and this was independent of smoking history.34 While the bimodal model remains a possibility, the first RCT of CT screening (NLST) has probably demonstrated some stage-shift effect and a change in mortality, although the follow-up period in this study is still relatively short. More data is needed.

BENIGN NCN

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

There is no doubt that the very high numbers of NCN detected by CT screening have lead to a major effort to learn how to deal with these. Most studies have reported very high rates of nodule detection, for example, in the Mayo Clinic study, in 69% of subjects at 2 years.35 In the NELSON trial, 8623 nodules of any size were detected at baseline screening of 7557 subjects,14 and in the NLST, approximately 24% of baseline screens were positive, of which 96.4% were false positive results. What matters is the identification of true cancers and to be able to distinguish these from benign NCN. The approach has been refined over time but is complex and beyond the remit of this review. Suffice it to say that having originally followed the Fleischner Society guidelines based on nodule diameter and then in incidence screens on growth, algorithms for intervention have also been developed. For example, in the New York ELCAP, 14% of 1000 screened subjects had suspicious findings, and 93% of invasive work-ups were positive for cancer.36 The NELSON trial has introduced volumetric analysis of NCN and classification of risk based on size, with nodules <50 mm3 regarded as benign, those 50–500 mm3 as indeterminate, and those >500 mm3 as positive and requiring investigation. This strategy has resulted in 94% diagnostic sensitivity and a 99% specificity at baseline screening.14 The UKLS single-screening study that has recently commenced will use a refined model based on the NELSON data to identify suspicious nodules, and either refer these patients for investigation or follow them up by further scanning, in an effort to further improve screening sensitivity and avoid further annual examinations in subjects with normal baseline scans.13 It is also worth noting that the application of minimally invasive surgery to nodules detected by screening has reduced perioperative complications, with the results being comparable with those for resection of cancers by lobectomy. The operative mortality for cancers detected by screening is 0.08% compared with 3.8% for conventional resection of early-stage cancers detected by standard methods.37

IS THERE A HEALTHY VOLUNTEER EFFECT?

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

A potential problem with a positive screening result from the NLST is that the ‘wrong’ subjects will volunteer for screening. There is already a commercial market for screening CT scans, often as a gift to partners who reach the age of 50 or 60 years, and these CT scans have no proven value. Furthermore, as discussed previously, the high-risk smoker is likely to be more reluctant to volunteer for CT screening.

It is well known that subjects who volunteer for screening trials are in general younger, healthier, better educated and smoke less than the average population. This was recently demonstrated in the large Prostate, Lung, Colorectal and Ovarian Cancer screening trial.38 While the standardized mortality ratio for all cause mortality in the population was 100, it was 31 at entry and increased to 48 at year 7. The effect of this factor on lung cancer RCT is unknown, although the population participating in the NLST is deemed to be representative of the relevant American population.

COST-EFFECTIVENESS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

This is a very important issue for health budgets. The cost-effectiveness of the NLST is not yet known but will be the subject of an important publication. Previous studies assessing the cost of screening for lung cancer have produced conflicting data and costs. Estimates have varied from $2500 to $2.3 million per life-year saved. Mahadevia et al.39 used a computer program to compare CT screening with no CT screening for 100 000 hypothetical subjects. Aiming for a stage shift of 50%, they estimated a cost of $116 300 per quality adjusted life year (QALY) gained for current smokers and $2.3 million per QALY gained for ex-smokers. Chirikos et al.40 estimated the economic value of screening two hypothetical cohorts. They concluded that, if 50% of lung cancers were detected when at a localized stage, a QALY would cost $48 000. The lowest estimate was from the ELCAP trial in which it was calculated by Wisnivesky et al.41 that a single baseline low-dose CT screen cost $2500 per life-year saved. However, the cost of a baseline screen would be >$50 000 per year if overdiagnosis bias exceeded 50%. The first real, large-scale prospective data will come from the NLST.

RADIATION EXPOSURE

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

An important criticism levelled at CT screening is the potential long-term increase in the relative risk of carcinogenesis due to radiation exposure. It is known that radiation exposure and smoking may have additive effects on lung cancer risk,42 and the risk of breast cancer may also be increased in patients undergoing serial cross-sectional imaging. Cody and colleagues compiled radiation-dose measurements from all 96 CT scanners used at the 33 centres participating in the NLST.43 They found significant heterogeneity in the radiation doses between scanners and also between CT scanner models from a single manufacturer. The more advanced 14- and 64-detector scanners showed greater dose efficiencies. Brenner44 modelled the potential increase in the number of thoracic malignancies due to lung cancer screening and concluded that the incremental lifetime lung cancer risk from a single low-dose CT scan was small. However, if CT screening was repeated annually for 50% of the eligible population between the ages of 50 and 75, there may be a 1.8% increase in the number of lung cancers. Ultimately, this issue will be definitively clarified only by long-term follow-up of patients participating in randomized CT-screening trials.

CONCLUSIONS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES

Screening for lung cancer using low-dose CT is a promising intervention provided that it is directed at an appropriate high-risk population that is still fit enough to undergo resection if a tumour is detected. The single report from the NLST RCT indicated that it is the most positive trial of lung cancer screening to date. Findings from other smaller RCT are due to be reported, and these will add weight to the existing data. However, many pitfalls remain to be overcome, as detailed in this review, before the success of CT screening can be confirmed.

At the same time, other methods of early detection, such as the measurement of volatile organic compounds in breath, and sputum cytology in association with CT or bronchoscopy should not be abandoned, and trials assessing these tools need to be completed. There is also on-going research to detect peripheral blood markers that can be used to identify subjects who are at high risk of lung cancer; once identified, these individuals may be followed up by CT.

It is too early to know whether low-dose CT is a cost-effective screening method, although it does appear to result in a reduction in mortality due to lung cancer. Nevertheless, most tumours detected by CT are around 15 mm in diameter and are detected when a long way down their natural pathway of progression or two thirds of the way through their life span, that is, late in their natural course. Much more information is needed, but the current revival of enthusiasm for lung cancer screening seems eminently justified.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. LOW-DOSE CT SCREENING FOR LUNG CANCER
  5. BIAS
  6. STAGE SHIFT
  7. CHOICE OF AT-RISK POPULATION
  8. SMOKING CESSATION PROGRAMS
  9. THE ACCEPTABILITY OF SCREENING
  10. ARE CANCERS DETECTED BY SCREENING DIFFERENT FROM OTHER CANCERS?
  11. BENIGN NCN
  12. IS THERE A HEALTHY VOLUNTEER EFFECT?
  13. COST-EFFECTIVENESS
  14. RADIATION EXPOSURE
  15. CONCLUSIONS
  16. REFERENCES