7
Adopting New Technology in the Face of Uncertain Science: The Case of Screening for Lung Cancer1

Lungcancer, anuncommontype ofcancer atthestartof the20th century, is the leading cause of cancer death in the United States at the start of the 21st. Surpassing deaths from breast, colon, and prostate cancer combined, there were an estimated 155,000 deaths from lung cancer in 2002 (ACS, 2002a). The prognosis after diagnosis is dismal. Five-year sur-vival rates remain less than 15 percent, changing little over the past 30 years (Travis et al., 1995). While lung cancer is mostly preventable through avoidance of tobacco products, smokers, health care providers, and scientists have unsuccessfully tried other preventive approaches, such as screening for early disease with chest radiographs and sputum cytology (secondary prevention).

Finding cancer earlier by screening seems intuitively appealing. Successful early detection of cervical cancer with Pap testing, breast cancer with mammography, and colon cancer through finding and removing polyps has lowered mortality from these cancers, thus providing impetus to search for early detection methods for other cancers. Unfortunately, the value of screening for other cancers, such as prostate-specific antigen (PSA) testing for prostate cancer, is less clear and has become more contentious (see also Chapter 5). Some cancers may be more amenable to early detection methods than others. For lung cancer, the prominent failures of chest radiographic and sputum cytology screening to lower disease mortality have led most organizations to recommend against screening for it.

1  

This chapter is based on a background paper prepared by Parthiv J. Mahadevia, Farin Kamangar, and Jonathan M. Samet (www.iom.edu/ncpb).



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection 7 Adopting New Technology in the Face of Uncertain Science: The Case of Screening for Lung Cancer1 Lungcancer, anuncommontype ofcancer atthestartof the20th century, is the leading cause of cancer death in the United States at the start of the 21st. Surpassing deaths from breast, colon, and prostate cancer combined, there were an estimated 155,000 deaths from lung cancer in 2002 (ACS, 2002a). The prognosis after diagnosis is dismal. Five-year sur-vival rates remain less than 15 percent, changing little over the past 30 years (Travis et al., 1995). While lung cancer is mostly preventable through avoidance of tobacco products, smokers, health care providers, and scientists have unsuccessfully tried other preventive approaches, such as screening for early disease with chest radiographs and sputum cytology (secondary prevention). Finding cancer earlier by screening seems intuitively appealing. Successful early detection of cervical cancer with Pap testing, breast cancer with mammography, and colon cancer through finding and removing polyps has lowered mortality from these cancers, thus providing impetus to search for early detection methods for other cancers. Unfortunately, the value of screening for other cancers, such as prostate-specific antigen (PSA) testing for prostate cancer, is less clear and has become more contentious (see also Chapter 5). Some cancers may be more amenable to early detection methods than others. For lung cancer, the prominent failures of chest radiographic and sputum cytology screening to lower disease mortality have led most organizations to recommend against screening for it. 1   This chapter is based on a background paper prepared by Parthiv J. Mahadevia, Farin Kamangar, and Jonathan M. Samet (www.iom.edu/ncpb).

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection SOURCE: Corbis Corporation. Recently a “high-tech” medical imaging device called spiral or helical computed tomography (CT) scan has renewed hope for finding an early detection method that can reduce mortality from lung cancer (Brice, 2000). Promising preliminary studies report that spiral CT scans can detect lung cancers at a smaller size than can chest radiographs (Henschke et al., 1999; Henschke et al., 2001; Sobue et al., 2002; Sone et al., 2001; Swensen et al., 2002). However, the clinical significance of these findings is unclear since long-term outcome data are unavailable. Randomized controlled trials evaluating spiral CT screening for lung cancer have only recently begun and conclusive efficacy data could be 5 to 10 years away. Despite the lack of clear benefit, direct-to-consumer marketing of spiral CT screening is being offered by entrepreneurial radiology practices (Lee and Brennan, 2002). Early dissemination of an unproven screening test raises many concerns and questions. Concerns include false-positive and false-negative tests, harms from subsequent invasive procedures or treatments, and sizable costs to consumers, payers, and society. Decision makers have many questions. Should consumers get these scans? How should health care providers counsel high-risk individuals interested in this technology? Should managed care organizations and other third-party payers cover the costs of screening? What experimental or observational study designs provide the best data in the most efficient manner?

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection The case study presented in this chapter evaluates this high-technology screening test through a review of past and current scientific evidence. Clinical studies of lung cancer screening techniques have close to a 50-year history. Using a historical perspective, we review the lessons learned from past attempts to assist individuals, clinicians and policy makers in making decisions on the use of lung cancer screening technology despite the uncertainty of its effectiveness. SCREENING FOR LUNG CANCER BY CHEST RADIOGRAPHY AND SPUTUM CYTOLOGY In the early 1950s several researchers noted that “X-ray surveying” of the population detected lung cancers in asymptomatic individuals (Lilienfeld, 1966), raising the possibility that screening for lung cancer by chest radiography might detect cancers at earlier stages, when there might be hope of operative resection and cure. At that time there was already lengthy experience with mass screening for tuberculosis with a similar goal: identification of cases at a stage when intervention was most likely to be effective. Screening for tuberculosis was a major public health activity, and screening clinics with mobile radiographic facilities were successfully used for this purpose. It seemed reasonable to extend these same approaches to an emerging epidemic of another fatal pulmonary disease. Four prospective cohort studies of lung cancer screening were started in the 1950s to determine if screening by chest radiography could improve lung cancer survival rates: the Veterans Administration-American Cancer Society (VA) Study (Lilienfeld, 1966), the Philadelphia Neoplasm Research Project Study (Weiss et al., 1982), the South London Lung Cancer Study (Nash et al., 1968), and the Tokyo Metropolitan Government Study (Hayata et al., 1982). Those studies used survival data to evaluate effectiveness and found 5-year survival rates that ranged from 8 to 20 percent (Table 7.1), not a meaningful improvement from the historical lung cancer survival rate. Survival rates among patients who had undergone surgical resection were 12 to 44 percent, higher than the overall survival rate. Unfortunately, the four studies did not incorporate control groups, and any improvement in survival from screening could not be assessed. Two other nonrandomized studies, the North London Lung Cancer study (Brett, 1969) and the Erfurt County, Germany, study (Wilde, 1989), evaluated screening by chest radiography and did have control groups (Table 7.1). Both studies included an intervention group that received chest radiographs every 6 months and a control group that had either no screening or less frequent screening than the intervention group. In both studies, the 5-year survival rate was higher among the intervention group than the control group (15 versus 6 percent in the North London Lung Cancer study and 14 versus 8 percent in the Erfurt County study). However, lung cancer

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection TABLE 7.1 Summary of Nonrandomized Prospective Trials of Lung Cancer Screening (1950s to 1970s) Study Veterans Administration American Cancer Society Study, 1958–1961 (Lilienfeld, 1966) Philadelphia Neoplasm Research Project, 1951–1965 (Weiss et al., 1982) Design Population and Number Screeneda Uncontrolled prospective study 14,607 males ages 45 and older Uncontrolled prospective study 6,136 males ages 45 and older Screening Interval and Method 6-month chest radiographs and sputum cytology 6-month chest radiographs Incidence Rate (per 1,000 person-years) 0.52 percentb 2.3 percent Number of Cancers Found 73 cases 121 cases Overall 5-Year Survival Rate 17 percentc 8 percent Percentage of Cancers Resected 36 percent 27 percent 5-Year Survival Among Those Who Had Resection 12 percent 18 percent Lung Cancer Mortality Rate (per 1,000 person-years) 0.7 percentb 47 percent Number of Cancers Found Between Screenings 5 cases NRd Comments High attrition rate; VA domiciliary sample Volunteer sampling; high attrition rate aNumbers are incidence screened. bReported as a proportion (percent) of all patients only rather than as a rate. cReported as the 32-month survival rate. dNR = not reported. eReported as 4-year survival rate.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection Tokyo Metropolitan Government study (Hayata et al., 1982) South London Lung Cancer Study, 1959–1963 (Nash et al., 1968) North London Cancer Study, 1959, (Brett, 1969) Erfurt County Study, 1953–1979 (Wilde, 1989) Uncontrolled prospective study 1,871,374 radiographs Uncontrolled prospective study 67,400 males ages 45 and older Controlled prospective study Screened group, 29,733; control group, 25,311 males ages 40 and older Controlled prospective study Screened group, 41,532; control group, 102,348 males Annual chest radiographs 6-month chest radiographs 6-month chest radiographs Screened group, 6-month chest radiographs; control group, 18-month chest radiographs 10.3 cases/100,000 radiographs 1.4 percent Screened group, 1.1 cases; control group, 1.0 case Screened group, 0.9 percent; control group, 0.65 percent 193 cases 147 cases Screened group, 101 cases; control group, 77 cases Screened group, 374 cases; control group, 667 cases 20.6 percent 27 percente Screened group, 15 percent; control group, 6 percent Screened group, 14 percent; control group, 8 percent 56 percent 56 percent Screened group, 44 percent; control group, 29 percent Screened group, 28 percent; control group, 19 percent 43.6 percent 47 percente Screened group, 32 percent control group, 23 percent Screened group vs. control group, 52 vs. 27 percent; 10-year: 39 percent vs. 19 percent NR NR Screened group, 0.7 percent; control group, 0.8 percent Screened group, 0.8 percent; control group, 0.6 percent 67 cases 87 cases (estimated) 36 cases Screened group, 199 cases; control group, 485 cases No mention of attrition or compliance High attrition rate High attrition rate County-specific study

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection mortality rates were the same in both the intervention and the control groups. The discrepancy between the improved survival rate and the unchanged mortality rate was later explained by the previously mentioned biases that often affect screening data (see also Chapter 5). These early studies were nonrandomized intervention studies that might be called “demonstration projects” today. Although the clinical trial was an established method for the evaluation of therapeutic interventions at the time, it had not yet been applied to the evaluation of screening. The landmark randomized controlled screening trial—the Health Insurance Plan of New York, which studied breast cancer—was not started until the mid-1960s (Shapiro, 1997). Although most of the early lung cancer screening studies used chest radiography as the principal screening test, the VA study also evaluated sputum samples as another method for the detection of cancer. Oscar Auerbach (Auerbach, 1969), a pathologist, showed that a spectrum of histologic abnormalities could be found in the respiratory epithelia of smokers, ranging from normal cells to frank malignancy. Geno Saccomanno and colleagues (Saccomanno et al., 1974), who developed the techniques needed for the preparation of specimens of respiratory cells for cytological examination, showed that this spectrum of abnormalities was mirrored in exfoliated cells from the lung. These observational studies provided a rationale for screening for lung cancer by cytological examination of sputum (sputum cytology), which was considered a screening technique complementary to chest radiography. Radiography was presumed to be better at finding radiographically visible peripheral cancers, which originate in the small airways and alveoli (air sacs) of the lung, whereas sputum cytology would find centrally located and hence radiographically invisible cancers arising from the larger airways of the lung, the bronchi. The VA study estimated that the addition of sputum cytology increased the rate of detection of lung cancer by 50 percent compared with that by the use of chest radiography alone. On retrospective assessment, these early lung cancer-screening studies had serious flaws, including a failure to have a control group and to randomize the participants to screened and nonscreened groups. Consequently, the results may have been affected by the time-related biases that arise in screening studies. Their results were also limited by attrition of the study populations, poor compliance with the screening regimen, difficulties with sputum collection, and high rates of mortality from surgery. For screening to be effective, most enrollees should be compliant with the screening regimen. In the VA study, roughly 30 percent of the initial enrollees failed to return for a second chest radiograph. By the third year, only 685 of the initial 14,607 enrollees received their recommended chest radiographs. The Philadelphia Neoplasm Research Project study noted that noncompliant individuals had a 76 percent higher rate of lung cancer than participants who complied. If dropouts are more likely to have the disease,

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection the effectiveness of any screening program may seem to be lower, as the individuals at greatest risk are less likely to receive the intervention. The quality of screening by sputum cytology in these studies was not optimal. International standards for cytological classification were not yet developed, and there was a high degree of variability in interpretation of abnormalities that fell between the normal and the malignant states (Fullmer, 1970). The significance of finding “atypical” cells was unclear. In 1970, Fullmer noted that priority areas for the enhancement of sputum cytology as a screening method included further refinements in the sample collection technique, education of the technicians who performed the cytological examination, reductions in costs, and establishment of international standards. A positive sputum cytology result requires follow-up by another test to localize the cancer. The VA study had difficulty finding the lung cancer when the sputum cytology result was considered positive but the chest radiography result was negative. The poor localization of cancer made surgical resection less effective, if not impossible. Finally, postoperative death rates were high, approaching 30 percent in the Philadelphia Neoplasm Research Project study. Randomized Controlled Trials of Chest Radiographic and Sputum Cytology Screening Building on the earlier studies, the NCI sponsored three randomized controlled trials of lung cancer screening in the 1970s: the Johns Hopkins Lung Project (the Hopkins study) (Tockman, 1986), the Memorial Sloan Kettering Lung Project (the Memorial study) (Melamed et al., 1984), and the Mayo Lung Project (the Mayo study) (Fontana et al., 1986). A fourth randomized controlled trial was performed in the Czech Republic (Kubik and Haerting, 1990). The studies addressed the key design deficiencies of the earlier studies: assignment to screening was by randomization, and careful conduct of the studies addressed issues of compliance and attrition (Berlin et al., 1984). Technological advancements such as CT and flexible fiberoptic bronchoscopy improved the ability to localize the cancer in persons positive by screening. In addition, surgical techniques had improved, and the postoperative mortality rate had declined since the earlier studies. The individuals in the intervention arms of all three NCI studies underwent both chest radiography and sputum cytology every 4 months. The individuals in the control arms of the Hopkins and Memorial studies also underwent chest radiography annually. The Mayo study had a different control arm; enrollees were given advice only at the time of enrollment to have chest radiography and sputum cytology performed annually. In the Czech study, which lasted 3 years, the intervention group underwent chest radiography and sputum cytology every 6 months, whereas the control group had both tests at the beginning and at the end of the study.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection TABLE 7.2 First Screening (Prevalence) Results from NCI-Sponsored Randomized Controlled Trials of Lung Cancer Screening Using Chest Radiographs and Sputum Cytology Study Johns Hopkins Lung Project, 1973–1978 (Tockman, 1986) Memorial Sloan Kettering Lung Project, 1974–1978 (Melamed et al., 1984) Population and Numbers Screened 10,387 male volunteers, ages 45+ with median 28.5 pack-year history of smoking 10,040 male volunteers, ages 45+ with median 31.2 pack-year history of smoking Screening Intervention (number of subjects in each group) I (5,226): CXR and SC C (5,161): CXR I (4,968): CXR and SC C (5,072): CXR Prevalence Rate (per 1,000 persons) Overall: 7.6 I: 7.5 C: 7.8 Overall: 5.3 I: 6.0 C: 4.5 Number of Cancers Detected I: 39 C: 40 I: 30 C: 23 5-Year Survival Rate in Study Group(s) I: 59 percent C: 35 percent I: 47 percent C: 31 percent Number of Stage 1 Cancers Detected I: 26 C: 16 I: 14 C: 8 5-Year Survival Rate for All Stage 1 Diseaseb 90 percent 85 percent Percentage of All Cancers Resected I: 69 percent C: 42 percent I: 60 percent C: 48 percent Cancers Detected by Sputum Cytology Alone 11 9 Number of Second Primary Lung Cancers 8 6 Comments 2 postoperative deaths 2 postoperative deaths; 16 surgeries for non-malignant lesions NOTE: Results are reported separately for groups receiving the intervention (I) and those that were controls (C). CXR = chest radiography; SC = sputum cytology. aThe NCI intervention group includes all of the Mayo study subjects and the intervention groups in the Hopkins and Memorial studies. These are then compared with the control groups in the Hopkins and Memorial studies.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection Mayo Lung Project, 1971–1976 (Fontana et al., 1986) NCI composite results of the above three trialsa (Berlin et al., 1984) Czech study (Kubik and Haerting, 1990) 10,933 male volunteers, ages 45+ with median 20 pack-year history of smoking 31,360 males 6,364 males ages 40–64 with 32-year smoking history All enrollees received CXR and Sputum Cytology I (21,127): CXR and SC C (10,233): the control cases in the Hopkins and Memorial studies All enrollees received CXR and Sputum cytology Overall: 8.3 Overall: 7.1 I: 7.6 C: 6.2 Overall: 3.0 Overall: 91 I: 160 C: 63 Overall: 19 Overall: 40 percent Hopkins/Memorial I: 55 percent; Mayo group: 40 percent; Hopkins/Memorial C: 35 percent Overall: 45 percent Overall: 26 percent Overall: 41 Overall: 105 I: 81 C: 24 Overall: 5 70 percent 80 percent NR 54 percent I: 76 percent C: NRc 33 percent 17 37 NR 7 21 NR 3 postoperative deaths; 28 surgeries for non-malignant lesions   bThese survival rates reflect those that were resected, not all stage I disease. cNR = not reported.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection TABLE 7.3 Incidence Screening Results from Randomized Controlled Trials of Lung Cancer Screening Using Chest Radiographs and Sputum Cytology Study Johns Hopkins Lung Project, 1973–1978 (Tockman, 1986) Population and Numbers Screened 10,387 male volunteers ages 45+ with median 28.5 pack year history of smoking Screening Intervention (numbers in each group I (5,226): CXR and SC C (5,161): CXR Incidence rate (per 1,000 person-years) I: 4.6b C: 4.9b Number of Cancers Detected I: 155 C: 162 5-Year Survival in Study Groups I: 20 percentc C: 20 percentc Number of Early vs. Advanced Cancers Founda I: early vs. advanced 83 and 111 C: early vs. advanced 93 and 109 Percentage of All Cancers Resected I: 47 percentc C: 44 percentc 5-Year Survival for Cancers That Were Resected NRd Mortality Rate (1,000 person-years) I: 3.4 C: 3.8 Number of Cancers Found Between Screenings or Due to Symptoms 193 total Additional Number of Cancers Found by SC 22 NOTE: Results are reported separately for groups receiving intervention (I) and those that were controls (C). CXR = chest radiography; SC = sputum cytology. aEarly cancers are those staged as 0, 1, or 2, and late cancers are those staged as 3 or 4. bThese results are based on interim results; the final results did not report these statistics. cEight-year survival. dNR = not reported. eThese results are for stage 1 cancers only.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection Memorial Sloan Kettering Lung Project, 1974–1982 (Melamed et al., 1984) Mayo Lung Project, 1971–1976 (Fontana et al., 1986) Czech study (Kubik and Haerting, 1990) 10,040 male volunteers ages 45+ with median 31.2 pack-year history of smoking I (5,072): CYR and SC C (4,968): CXR 10,933 male volunteers ages 45+ with median 20 pack-year history of smoking I (4,618): CXR and SC C (4,593): annual advice to get CXR 6,364 males ages 40–64 with 32-year smoking history I (3,171): CXR and SC every 6 months for 3 years C (3,174): CXR and SC 3 years apart I: 3.7 I: 5.5 I: 6.0 C: 3.8 C: 4.3 C: 4.5 I: 114 I: 206 I: 108 C: 121 C: 160 C: 82 I: 36 percent I: 33 percent I: 18 percent C: 33 percent C: 15 percent C: 18 percent I: early vs. advanced, 54 and 85 I: early vs. advanced, 99 and 107 I: early vs. advanced, 55 and 53 C: early vs. advanced, 68 and 86 C: early vs. advanced, 51 and 109 C: early vs. advanced, 36 and 46 I: 51 percent I: 46 percent I: 23 percent C: 53 percent C: 32 percent C: 23 percent 80 percente 50 percent 26 percent I: 2.7 I: 3.2 I: 3.6 C: 2.7 C: 3.0 C: 2.6 I: 44 I: 116 I: 47 C: 56 C: 160 C: 44 18 18 2

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection ing trials. For example, in the Shinshu University study, 1,035 of 5,460 participants without identifiable cancer, 19 percent of all eligible participants, did not return for their second year screening. In contrast, the Mayo Clinic Study reported a low non-compliance rate of only 3 percent per year, suggesting that low rates of non-compliance are achievable. One study, ALCA, has reported 5-year survival estimates. Among lung cancers found with prevalence screening, the 5-year survival was 76.2 percent and among those found on incidence screening, the 5-year survival was lower, 64.9 percent. The authors acknowledged that these estimates could be influenced by length, lead-time, and overdiagnosis biases. The lower survival among lung cancer participants detected during incidence screenings suggests the presence of length bias among the prevalent screening cancers. Implications of Lung Cancer Screening with Spiral CT Scans The results from these uncontrolled trials are promising and leave little doubt that spiral CT is more sensitive in detecting lung cancers than chest radiographs. However, while encouraging, they do not provide conclusive evidence for long-term efficacy. Even had they provided long-term outcome data, a full understanding of the clinical significance of the results would not have been gained as these trials, like the demonstration projects of the 1950s, all lack a control group. Simply finding smaller-sized cancers does not mean mortality is lowered. In one study, among localized stage cancers, smaller tumor size was not associated with better outcomes, demonstrating that some biologically aggressive forms of lung cancer may metastasize early, even when they are 1–5 mm in size (Patz et al., 2000b). Elevated survival rates also do not prove efficacy. Survival rates are affected by selection, lead-time, length and over-diagnosis bias (see Chapter 5 for definitions), hence inferences about screening efficacy made from these data alone become speculative. Cancers with longer latency periods or with the potential for length bias are likely to be over-sampled in the early screening years, and inflated survival estimates may result, as shown in the ALCA results. Overdiagnosis bias is also a concern, since very small cancers have been found and the natural course of disease in individuals with tumors of this size is not known. Could some of these cancers grow very slowly, fail to progress, or perhaps regress? Excluding the possibility of overdiagnosis bias will be difficult, as the identification of a lung cancer mandates curative therapy and observation without therapy would not be considered ethical. Estimates of overdiagnosis bias may best be gained after the fact through autopsy screening (Black, 2000). Despite the many weaknesses of the CT studies, they have spurred interest and investments into randomized controlled trials of spiral CT screening.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection Concerns such as physical and psychological harms from screening have been raised in the spiral CT screening studies. A case of a 73-year-old woman, Mrs. S., who decided to undergo lung cancer screening with spiral CT, illustrates the downside of screening. As described in the New York Times, Mrs. S.’s spiral CT scan showed a collapsed lung, presumably due to an obstructing lung cancer. However, after open lung surgery, surgical pathology showed no evidence of lung cancer. Her doctor writes, “While this news is welcome, Mrs. S.’s surgery and a rocky postoperative course had drained her both physically and emotionally. When she returned home, it took her several months to recover. She is still paying her hospital bills” (Lerner, 2002, p. D6). This individual did not benefit from screening and her case describes the possibility of unintended consequences of spiral CT screening. Psychological harms can affect many individuals in these trials. Uncertainty as to the diagnosis of an indeterminate lung nodule can cause much anxiety, as the affected individuals may have had to wait months to years before learning that their nodules were not growing and hence not of concern. More than half of all participants in the Mayo Clinic Study had indeterminate nodules. This study, which used the most updated screening technology, found a higher rate of lung nodules than in other studies. Trials of spiral CT face the difficult challenge of appropriately triaging these very common nodules and counseling participants. Wardle and Pope (1992) pointed out that psychological costs from early detection are worrisome, and Reich argued that so far lung cancer “screening does no good and may do much harm” (1995, p. 557). Edward Golub (1999) wrote, “It is not overdramatic to say that the entire nature of the future life of a patient can depend on the results of...[these] tests, what the person can and cannot do, how much time the person has to do it in, what the person’s self-perception is, and how others think of and behave toward the person” (p. 13). The implications for otherwise healthy participants of screening tests are even more striking since they are at risk of being “transformed” from wellness into sickness. Another potential harm of screening would occur if smokers receiving a negative screening test decided that they did not need to quit smoking. These spiral CT studies have been performed at referral centers with highly motivated staff, investigators, and participants. If widespread mass screening by spiral CT were adopted today, could their results be replicated elsewhere? It is possible that they could, but the costs of developing the infrastructure required to deliver top-quality care would be substantial. Standards for the detection and measurement of lesions by spiral CT scanning need to be established. Costs for the training of personnel and investments to purchase scanning instruments are required. Furthermore, capacity would also be needed to carry out diagnostic workups and follow-up. Given the potential harms, logistical hurdles to minimize harm and

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection costs, and, most importantly, the lack of evidence of efficacy, judgment on using spiral CT as a lung cancer screening test should be reserved until evidence from well-designed clinical trials can be evaluated. Sputum Cytology as an Adjunctive Screening Test While spiral CT scans are superior to chest radiographs for detecting lung cancer, CT could still miss cancers hiding in endobronchial locations. Sputum cytology is considered a good adjunctive screening test since it frequently detects endobronchial cancers, which are usually squamous cell carcinomas. The randomized controlled trials that were started in the 1970s showed that of the four major histologic types of lung cancer, the best prognosis was for squamous cell carcinoma. Squamous cell cancers represented 25 percent of all cancers found, and the 5-year survival rate for those with squamous cell cancers detected by cytology only was 85 to 90 percent (Berlin et al., 1984). Another cohort study that monitored lung cancer patients with radiographically occult malignancies identified by sputum cytology reported 5-year survival rates of 74 percent (Bechtel et al., 2000). Kennedy and colleagues (2000) pointed out that the number of deaths caused by squamous cell carcinoma of the lung is similar to the number caused by breast or colon cancer, for both of which screening is recommended. Thus, if squamous cell carcinoma is the slowest growing of the lung cancers and therefore the most likely to be detected early, screening for this particular type of lung cancer by sputum cytology may be warranted (Kennedy et al., 2000). Advances in the screening of sputum samples hold promise, as do new bronchoscopic methods for examination of the lung. Researchers have identified precancerous and early cancerous states in sputum by identification of certain abnormal genes in sputum cells and improved localization of early cancers through fluorescent bronchoscopy, which consists of the identification of malignant cells by bronchoscopic examination under fluorescent light (Lam et al., 1998; Palmisano et al., 2000; Tockman, 2000). For example, Palmisano and colleagues (2000) found that certain cancer-fight-ing genes in the sputum cells of smokers had an abnormality called hypermethylation. On examination of sputum specimens several years before cancer developed, they found that hypermethylation antedated lung cancer in each of 21 persons who eventually developed lung cancer. They reported that “aberrant methylation ... can be detected in DNA from sputum in 100 percent of patients with squamous cell lung carcinoma up to 3 years before clinical diagnosis” (Palmisano et al., 2000, p. 5954). This specific molecular abnormality is only one of many that appear to be promising as targets for early detection, and new bronchoscopic methods should eventually improve the ability to find very small tumors. Fluorescent bronchoscopy is more sensitive than the traditional means

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection of examination under white light for the detection of cancers. Kennedy and colleagues (2000) pointed out “the increased sensitivity [of fluorescent bronchoscopy] is associated with decreased specificity, resulting in many false positive biopsies” (p. 76S). A low specificity adds to the costs of using the fluorescent bronchoscope and increases the time of the procedure. Bias from overdiagnosis might also be introduced by the detection of very small cancers. Whether molecular or genetic markers in sputum, accompanied by bronchoscopic examination, can find curable lung cancers has yet to be shown. FUTURE DIRECTIONS IN LUNG CANCER SCREENING The promise of new technologies has led to the initiation of randomized controlled trials. Randomized studies of chest radiographs and spiral CT scanning for lung cancer screening are under way (Patz et al., 2000a). The Prostate, Lung, Colorectal, and Ovarian (PLCO) study has randomized 152,000 participants to receive screening chest radiographs or no screening (Simpson et al., 2000). The National Lung Screening Study, using a subset of the PLCO, will randomize 50,000 participants to chest radiographs or spiral CT screening. This NCI-sponsored trial is collaborating with the American College of Radiology Imaging Network (ACRIN) to enroll heavy smokers between the ages of 55 and 74. The trial will screen for 3 years with a 4-year follow-up, also collecting sputum samples in 10,000 participants. The trial will conclude data collection in 2009 (Sullivan, 2002). How long before conclusive effectiveness data will be published? The NCI chest radiographic screening trials started in the early 1970s, and reports were published in the mid-1980s. Conclusive data for spiral CT may not be available for at least 5 to 10 years. Japan, in contrast to the United States, has already adopted spiral CT for lung cancer screening, despite a lack of evidence supporting its effectiveness. Ecological population data on the effect of widespread mass screening on lung cancer incidence and mortality rates could provide clues as to whether such screening is effective, although this type of data, due its limitations, usually does not provide sufficient evidence to recommend screening. Already, techniques that are being used in practice are more advanced than those being considered in the most recent and ongoing studies. The initial spiral CT trials used single-detector CT technology, which is now outdated (National Cancer Institute and ACS, 2001). General Electric Medical Systems is selling multi-detector spiral CT scanners. Multi-detector scanners offer better resolution than prior machines and offer the option of magnifying parts of the lung without rescanning the participant. Within the next 5 to 10 years, even this technology will be updated (Fox, 2001). The NCI/ACRIN study, which is using multi-detector technology, is at risk of

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection reporting data based on technology that will be considered obsolescent before the trial is completed. These new scanners can detect 1- to 4-mm lesions called “ground glass opacities.” These lesions are so small that they cannot be characterized as nodules. By improving the sensitivity of scanning, specificity is likely to decline so that false-positive results are likely to increase. The Mayo Clinic used multi-detector technology and found a higher rate of lung nodules than trials using single detector technology. Clinicians will face the challenge of distinguishing between false-positive and true-positive results in order to prevent unnecessary morbidity and mortality. On a short-term basis, antibiotic therapy can be administered followed by repeat scanning to see if the opacity was of an infectious etiology. Software that provides computer-aided diagnosis (CAD) can be used to enhance the accuracy of reading. Software algorithms can estimate the likelihood of a malignancy on the basis of certain characteristics of the lesion and the participant. CAD software can also double-check the readings of radiologists and pathologists. PAPNET, a CAD device for the reading of Pap smears, has been shown to reduce the number of smears with false-negative results (Halford et al., 1999). Similar software is being developed for mammograms and CT scans (National Cancer Institute and ACS, 2001). CT imaging technology is also being used more widely. Virtual or three-dimensional means of bronchoscopy and colonoscopy imaging are being evaluated as noninvasive alternatives to conventional endoscopy (Black, 1999a). Spiral CT angiography evaluates arteriosclerosis without placing the individual at risk from the use of the contrast dye that is required by conventional angiography (Siegel and Evens, 1999). Spiral CT angiography has incidentally found lung cancers. Some radiology sites are offering screening by multiphasic imaging for cancer and arteriosclerosis (National Cancer Institute and ACS, 2001). If screening is not targeted, there is a high likelihood of finding many false-positive lesions. Public and Policy Reactions to New Technologies When the initial findings from the ELCAP study were reported in medical journals, major newspapers and weekly newsmagazines published articles about the findings. The articles mentioned that lung cancer death rates could be greatly reduced if smokers and former smokers were routinely given a CT test that can detect tumors when they are small enough to be cured (Brice, 2000). The public response to the news was dramatic. The voice mail systems at the hospital centers participating in the ELCAP study were overwhelmed, and at least 3,000 calls were placed to the Mayo Clinic by the next day. One ex-smoker whose husband died from lung cancer stated “I’m so grateful that the technology is present ... lung cancer is just an awful, awful thing.”

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection The public has long been worried about lung cancer. In an American Cancer Society survey that asked participants to “list the body sites susceptible to cancer that first come to mind,” respondents most commonly mentioned the lung, breast, and skin (ACS, 1980b, p. 93). “About seven out of 10 smokers (71 percent) believe that if lung cancer is detected early, there is a good chance that it can be cured” (ACS, 1980b, p. 98). As lung cancer is a dreaded disease and screening is viewed as bringing hope for its detection and cure, it should come as no surprise that the ELCAP study results were enthusiastically received. Some community physicians routinely perform screening chest radiographs, despite their lack of effectiveness as determined in clinical trials, and despite the lack of endorsement of their use by policy organizations (Black, 1999a). Adoption of CT technology by some providers is likely given their prior beliefs and practice patterns. Radiology groups have adopted this technology and are advertising their use of the technology in local newspapers and on television and radio (Brice, 2000). One group in Skokie, Illinois,, found one cancer in 120 screening procedures. The average price per scan was $325, all of which was paid out of pocket by the screening participant. The premature promotion of spiral CT as a lung cancer screening tool raises many questions. Should spiral CT be promoted even though its effectiveness is not established? Does the public understand the consequences of having this screening test? How well informed is the public concerning the unwanted consequences of false-positive test results? What are the conflicts of interest when providers who could make financial gains by screening advertise on the promise of an unproven technology? These reports suggest fast and too early adoption of this unproven technology. Cautions have been raised about spiral CT screening for lung cancer. Christine D. Berg of NCI’s Division of Cancer Prevention and Suburban Cancer Hospital Center reported that “We had leeches in the 1800s, radium elixirs in the 1900s and radiation treatment for enlarged thyroids in the 1950s. A long list of medical fads have come and gone. Spiral CT has great promise. That’s why it deserves further study” (Brice, 2000, p. 49). The Society of Thoracic Radiology, which included researchers of the Mayo trial and the ELCAP study, issued the following consensus statement: “It is the consensus of this committee that mass screening for lung cancer with CT is not currently advocated. Suitable subjects who wish to participate should be encouraged to do so in controlled trials, so that the value of CT screening can be ascertained as soon as possible” (Aberle et al., 2001, p. 65). The American College of Radiology has appointed a task force to evaluate spiral CT and withholds recommendation at the time of this writing (Zinninger MD, American College of Radiology, personal communication, 2001). Although the initial results of evaluations of spiral CT scanning are

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection encouraging, it is now known that radiographic detection of asymptomatic lung cancer provides no assurance that benefit in terms of the prevention of mortality from lung cancer will ensue. The same level of evidence that was available for chest X-ray and sputum cytology decades ago is now available for spiral CT, and any judgment on spiral CT as a screening modality should await the findings of trials and the availability of mortality rate data. Therefore, at this time, the evidence does not support a recommendation for either widespread screening or the screening of selective high-risk groups for lung cancer. This conclusion is based on the historical record and its lessons, the potential for harm from widespread screening, and the lack of proven effectiveness from current evidence on spiral CT. Given that spiral CT is already in use, what specific policy recommendations can be implemented now? Public education, mutual decision-making, and a focus on primary prevention are needed to aid consumers, providers, researchers, payers, and policy makers. Consumers Smokers, current or former, are at far greater risk of lung cancer than those who have never smoked. However, the absolute risk of developing lung cancer among even the at-risk smoker groups is relatively low. SEER Program data show that the incidence rate of lung cancer among those age 65 and older is about 3 to 6 per 1,000 persons per year. Given the relatively low incidence rates, any decision to be screened for lung cancer by spiral CT should be made with full consideration of the rates of true-positive versus false-positive results and the attendant benefits and risks. Knowledgeable providers must adequately communicate these complexities so that patients can weigh the risks and benefits of screening. Educational materials that clearly explain risk can assist with informed decision-making. Consumers who choose to undergo screening by spiral CT, despite the current lack of evidence, should be informed of the chance of finding true disease and the attendant risks of screening. Topics that should be covered include the costs of initial and follow-up tests, the potential physical harms from radiation and invasive secondary testing and surgery, and the potential psychological harms from misdiagnosis or mislabeling. An informed-consent form detailing these facts should be administered before the scan is performed. Consumers should be informed of the possibility that screening may result in their knowing of cancer early without lowering risk of death. In other words, they could be living longer with a diagnosis but not necessarily really living longer. Examples from past screening studies can be used to illustrate this point.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection Providers Current clinical practice guidelines advise against routine lung cancer screening. Neither chest radiographic nor spiral CT screening should be offered on a routine basis. Preferably, a primary care provider or specialty physician familiar with lung cancer screening and its likely limitations should appropriately counsel patients who desire testing. Consumers who lack primary care or who bypass their primary care provider and go directly to radiology groups should receive appropriate counseling before undergoing the procedure. Radiologists will need to arrange for appropriate followup care. If comorbid conditions such as heart disease or emphysema are present and could prevent eligibility for surgical resection, screening should be delayed until an appropriate evaluation is completed. Radiology groups endorsing or advertising their use of spiral CT scanning should acknowledge its experimental nature and should clearly state the current lack of evidence. Follow-up mechanisms for individuals with indeterminate results should be established so that noncompliance is minimized. To ensure the highest standards of care, professional organizations and provider groups should develop guidelines on the management of pulmonary nodules. Researchers Since there is substantial uncertainty about spiral CT’s effectiveness, resources should be made available for research. A randomized controlled trial will provide the clearest, most convincing evidence of possible effectiveness, and the level of evidence from trials with this design is considered the strongest for policy development. Others have called such trials unethical (Henschke and Yankelevitz, 2000), as this would be the case if spiral CT were clearly efficacious and such trials were simply testing a new or altered protocol. At present, however, randomized controlled trials of screening are ethically sound; the evidence about screening by spiral CT (or other modalities) has not yet been tipped to favor screening. Since observational studies are subject to selection bias and uncontrolled confounding, uncertainty often clouds observational evidence of effectiveness, leaving the question of whether the benefit was due to the intervention or to some unmeasured difference between groups. Observational studies are useful for evaluations of the natural history of disease, tumor doubling times, and biomarker analysis and should be done for these purposes. Neither study design—randomized or observational—would provide a result faster since mortality from the disease is the common and appropriate endpoint. While waiting for trial results, researchers could evaluate decision support models that simulate the natural history of disease and spiral CT. These models can identify important disease and test thresholds at which screening can be a viable option.

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection Additional research is needed to find high-risk participants most likely to benefit from lung cancer screening, find methods to improve compliance with cancer screening, risk stratify the numerous lung nodules that are found on screening, and prevent harms induced by screening. Policy Makers and Payers For coverage decisions, policy makers will want to know the cost-effectiveness of spiral CT. David Eddy (1981) estimated that primary prevention by smoking cessation is over 400 times more effective than screening by chest radiography. The direct costs to provide chest radiographic screening to 100,000 40-year-old smokers were estimated to be $500 million in 1980 dollars. Clearly, if resources are limited, primary prevention is more cost-effective. Given that the cost of spiral CT screening is higher than chest radiographs, the budgetary impact to implement systematic screening is likely to be much higher today. During periods of uncertainty, policies should be implemented that educate consumers, providers, and policy makers of the risks of early adoption. Widespread mass screening for lung cancer has not been endorsed by professional organizations, given the history of past attempts and the difficulties with present attempts. Rapid dissemination of screening technology before establishment of its effectiveness and prior to establishment of quality control standards could result in more harm than benefits. Currently, for every three cancers detected by spiral CT screening, two false-positive participants will undergo invasive procedures and potential subsequent harm. CONCLUSIONS Lung cancer is a dreaded disease with a difficult and frequently fatal course. It is now the leading cause of cancer death in the United States. Smoking prevention and cessation can prevent most cases. Yet, one-quarter of adults in the United States are still regular smokers, and youths continue to experiment with tobacco and often become addicted. Although rates of smoking have declined, continuation of the current lung cancer epidemic can be anticipated for decades to come. In addition, lung cancer death rates will soon surge in the developing world (Hoel et al., 1992; Pandey et al., 1999; Parkin et al., 1999). Logically, investigators have looked for approaches other than tobacco control to reduce the numbers of deaths from lung cancer. Screening was first attempted in the 1950s, soon after the scope of the lung cancer epidemic was recognized. The use of chest X-rays for screening appeared to be appropriate at the time and fit with the approach already in use for tuberculosis. Sputum cytology added another screening tool that also appeared

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection promising, as the abnormalities in expectorated cells mirrored the abnormalities in the respiratory epithelium where they originated. The first studies were observational rather than experimental in design. Although the randomized clinical trial is now the standard means for the assessment of screening tests, it had not yet been used for that purpose in the 1950s. The first studies of screening tests used a design equivalent to a demonstration project, inviting participants to have the screening test and then monitoring them over time. Viewed by today’s standards, those studies had poor quality control measures, inaccurate standardization and poor means of interpretation of test results, and high rates of noncompliance with the screening regimen by participants. The designs of the studies were also flawed because they did not include randomization to a control group and one or more screening modalities; in fact, comparison or control groups were lacking in several of the studies. In the 1970s four randomized controlled trials on lung cancer screening using chest radiography and sputum cytology were performed. Unfortunately, these randomized controlled trials of lung cancer screening showed no evidence of early detection of lung cancers by these tests (Fontana et al., 1991; Melamed et al., 1984; Tockman, 1986). The numbers of early-stage and late-stage cancers did not meaningfully differ between the study groups. Most disappointing was that on follow-up, the rates of mortality were the same among the screened and the unscreened participants. An effective screening test for lung cancer has not yet been identified, therefore organizations that develop screening guidelines do not recommend screening for it (Aberle et al., 2001; Biesalski et al., 1998; Eddy, 1980a; Eddy, 1980b). Although the studies of screening to date have failed to show mortality reduction, the lessons learned from these trials can pave the way for future research with newer technologies such as spiral CT scans. The conceptual framework for cancer screening illustrates its complexity and provides criteria for the evaluation of new technologies such as spiral CT. Future studies need to account for biases when measuring survival data. These biases are lead-time bias, length bias, and overdiagnosis bias (described in Chapter 5). Future studies not only must show evidence that screening can detect smaller cancers but also must show evidence that screening can produce stage shifts, including both more early cancers and fewer late cancers, followed by lower lung cancer mortality rates (Patz et al., 2000a). Potential concerns related to the psychological and physical harms from the finding of positive results and the management of patients with positive results need to be considered, as there is strong demand for screening by spiral CT and spiral CT has been adopted into practice, despite the uncertain and incomplete scientific evidence. Clinicians and the public are eager to have new approaches to the prevention of lung cancer; consequently, any new tool for prevention is

OCR for page 259
Fulfilling the Potential of Cancer Prevention and Early Detection likely to be hailed and perhaps not receive a sufficiently critical evaluation. The lessons learned from past experience with lung cancer screening call for restraint in the use of spiral CT screening. Findings from uncontrolled trials conducted to date do not provide the level of evidence required to endorse systematic screening with spiral CT scans.