Table 2. Methods and Quality of Controlled Trials of Lung Cancer Screening

Study Assembly of Comparable Groups: Randomization/Allocation Concealment Maintenance of Comparable Groups Outcomes Assessment: Validity of Method, Masking Attendance, Compliance, Contamination, Crossovers Analysis, Exclusions, and External Validity Study Quality
Northwest London Mass Radiography Service35,36 Cluster randomized by random number; examiners not clearly blind; comparable in age structure and smoking habits; no apparent occupational exposures 99% followup Cause of death determined from hospital records and General Register's office; blinding not described Intervention: 63%
Control: 63%
Cross-over: NR
Intention-to-treat analysis; no reported exclusions; age and smoking habits similar Fair
Kaiser Permanente33,34 Randomized by patient record numbers with concealed code; more chronic lung disease in intervention group (8.9% vs 7.5%) Poor followup Blind review of death Intervention:
60% underwent MHC
(mean, 6.8 exams)
Cross-over:
64% of controls had MHC
(mean, 2.8 exams)
Very low-risk population Poor
Memorial Sloan-Kettering37,63-67 Computer-generated randomization (not described); all cause mortality similar Formal protocol/algorithm for followup; 55 lost to followup All deaths reviewed by statisticians, clinicians, and pathologists blind to study group Dual screen:
63.2%
CXR:
65.2%
Intention-to-treat analysis; only exclusion was prior lung cancer Fair
Johns Hopkins38,68-72 Computer generated randomization (not described); allocation concealment unclear; fairly comparable when evaluated by age, smoking history, non-tobacco carcinogen exposure Formal algorithm for followup; 1.3% lost to followup All deaths reviewed by statisticians, clinicians, and pathologists blind to study group Uncertain; 19% withdrew from active screening Intention-to-treat analysis; formal protocol for evaluation; only exclusion was prior lung cancer Fair
Mayo Lung Project39,74-76,80 Randomization method not described; allocation concealment unclear; similar distribution age, smoking exposure to non-tobacco carcinogens, and pulmonary disease Adequate; good followup of all participants in both groups All deaths reviewed by statisticians, clinicians, and pathologists blind to study group; National Death Index used for latest followup Intervention:
75%
Cross-over:
73% of controls had CXR within last 2 yrs of study
Intention-to-treat analysis; formal protocol for evaluation; Mayo Clinic population with life expectancy estimates of 5 yrs Fair
Czech40,84,85 Randomization stratified by age, smoking history, socioeconomic status, residence, occupational exposure; allocation concealment unclear; no differences observed in these characteristics; all cause mortality, smoking-related deaths higher in intervention group Not well reported Cause of death ascertained from death certificates; autopsy in 1/3 of patients; blind review not described Intervention: 92.5%
Cross-over: rare
Significantly higher all-cause mortality in screened group, suggesting bias in randomization Poor
Wilde47 Nonrandomized; similar community distribution of smoking habits and economic structure; similar all-cause mortality rates; population age not described Adequate description; greater number dropouts in control group Blinding not described; nonsystematic ascertainment of cause of death NR Intention-to-treat analysis; no reported exclusions; mortality rates not adjusted for age Poor

Notes: CXR, chest x-ray; MHC, multiphasic health checkups; NR, not reported.

Return to Document